Title of Invention	MULTI-DIMENSIONAL NEIGHBORING BLOCK PREDICTION FOR VIDEO ENCODING
Abstract	The motion estimation techniques and video encoding device(s) described use a two dimensional pipeline to generate accurate motion estimation parameters for a current video block. The two dimensional pipeline uses previously calculated motion estimation parameters of relevant neighboring video blocks, including a preceding video block on a same row as the current video block, prior to the generation of the accurate motion vectors, motion vector predictors, and mode decision of the current video block The use of the two dimensional pipeline allows accurate motion vector prediction from neighboring video blocks previously not available, in the computation of motion vectors. Three engines may be used in the two dimensional pipeline, a fetch engine, an integer search engine and a fractional and spatial search engine. While the fetch engine and fraction and spatial search engine operate on one row, the integer search engine operates on another row.

Title of Invention

MULTI-DIMENSIONAL NEIGHBORING BLOCK PREDICTION FOR VIDEO ENCODING

Abstract

The motion estimation techniques and video encoding device(s) described use a two dimensional pipeline to generate accurate motion estimation parameters for a current video block. The two dimensional pipeline uses previously calculated motion estimation parameters of relevant neighboring video blocks, including a preceding video block on a same row as the current video block, prior to the generation of the accurate motion vectors, motion vector predictors, and mode decision of the current video block The use of the two dimensional pipeline allows accurate motion vector prediction from neighboring video blocks previously not available, in the computation of motion vectors. Three engines may be used in the two dimensional pipeline, a fetch engine, an integer search engine and a fractional and spatial search engine. While the fetch engine and fraction and spatial search engine operate on one row, the integer search engine operates on another row.

Full Text	FORM 2 THE PATENTS ACT, 1970 (39 of 1970) & THE PATENTS RULES, 2003 COMPLETE SPECIFICATION (See section 10, rule 13) "MULTI-DIMENSIONAL NEIGHBORING BLOCK PREDICTION FOR VIDEO ENCODING" QUALCOMM INCORPORATED, a company incorporated in the state of Delaware, U.S., of 5775 Morehouse Drive, San Diego, California 92121-1714, U.S.A The following specification particularly describes the invention and the manner in which it is to be performed. WO 2007/038357 PCT/US2006/037130 MULTI-DIMENSIONAL NEIGHBORING BLOCK PREDICTION FOR VIDEO ENCODING This present application claims the benefit of U.S. Provisional Application No. 60/719, 891, filed on September 22,2005, and assigned to the assignee hereof, which is hereby incorporated herein by reference in its entirety. TECHNICAL FIELD [0001] What is described herein relates to digital video processing and, more particularly, encoding of video sequences. BACKGROUND [0002] Digital video capabilities can be incorporated into a wide range of devices, including digital televisions", digital direct broadcast systems, wireless communication devices, personal digital assistants (PDAs), laptop computers, desktop computers, digital cameras, digital recording devices, cellular or satellite radio telephones, and the like. Digital video devices can provide significant improvements over conventional analog video systems in creating, modifying, transmitting, storing, recording and playing full motion video sequences. [0003] A number of different video encoding standards have been established for encoding digital video sequences. The Moving Picture Experts Group (MPEG), for example, has developed a number of standards including MPEG-1, MPEG-2 and MPEG-4. Other standards include the International Telecommunication Union (ITU) H.263 standard, QuickTime™ technology developed by Apple Computer of Cupertino California, Video for Windows™ developed by Microsoft Corporation of Redmond, Washington, Indeo™ developed by Intel Corporation, RealVideo™ from RealNetworks, Inc. of Seattle, Washington, and Cinepak™ developed by SuperMac, Inc. New standards continue to emerge and evolve, including the ITU H.264 standard and a number of proprietary standards. [0004] Many video encoding standards allow for improved transmission rates of video sequences by encoding data in a compressed fashion. Compression can reduce the overall amount of data that needs to be transmitted for effective transmission of video WO 2007/038357 PCT/US2006/037130 frames. Most video encoding standards, for example, utilize graphics and video compression techniques designed to facilitate video and image transmission over a narrower bandwidth than can be achieved without the compression. [0005] The MPEG standards and the ITU H.263 and ITU H.264 standards, for example, support video encoding techniques that utilize similarities between successive video frames, referred to as temporal or inter-frame correlation, to provide inter-frame compression. The inter-frame compression techniques exploit data redundancy across frames by converting pixel-based representations of video frames to motion representations. In addition, some video encoding techniques may utilize similarities within frames, referred to as spatial or intra-frame correlation, to further compress the video frames. [0006] In order to support compression, a digital video device includes an encoder for compressing digital video sequences, and a decoder for decompressing the digital video sequences. In many cases, the encoder and decoder form an integrated encoder/decoder (CODEC) that operates on blocks of pixels within frames that define the sequence of video images. In the International Telecommunication Union (ITU) H.264 standard, for example, the encoder typically divides a video frame to be transmitted into video blocks referred to as "macroblocks" (MB's) which may comprise 16 by 16 pixel arrays. The ITU H.264 standard supports 16 by 16 video blocks, 16 by 8 video blocks, 8 by 16 video blocks, 8 by 8 video blocks, 8 by 4 video blocks, 4 by 8 video blocks and 4 by 4 video blocks. Other standards may support differently sized video blocks. [0007] For each video block in the video frame, an encoder searches similarly sized video blocks of one or more immediately preceding video frames (or subsequent frames) to identify the most similar video block, referred to as the "best prediction block." The process of comparing a current video block to video blocks of other frames is generally referred to as motion estimation. Once a "best prediction block" is identified for a video block, the encoder can encode the differences between the current video block and the best prediction block. This process of encoding the differences between the current video block and the best prediction block includes a process referred to as motion compensation. Motion compensation comprises a process of creating a difference block, indicative of the differences between the current video block to be encoded and the best prediction block. Motion compensation usually refers to the WO 2007/038357 PCT/US2006/037130 act of fetching the best prediction block using a motion vector, and then subtracting the best prediction block from an input block to generate a difference block. [0008] After motion compensation has created the difference block, a series of additional encoding steps are typically performed to encode the difference block. These additional encoding steps may depend on the encoding standard being used. In MPEG4 compliant encoders, for example, the additional encoding steps may include an 8x8 discrete cosine transform, followed by scalar quantization, followed by a raster-to-zigzag reordering, followed by run-length encoding, followed by Huffman encoding. An encoded difference block can be transmitted along with a motion vector that indicates which video block from the previous frame (or subsequent frame) was used for the encoding. A decoder receives the motion vector and the encoded difference block, and decodes the received information to reconstruct the video sequences. [0009] It is highly desirable to simplify and improve the encoding process. To this end, a wide variety of encoding techniques have been developed. Because motion estimation is one of the most computationally intensive processes in video encoding, improvements to motion estimation can provide notable improvements in the video encoding process. [0010] It is desirable to find more efficient and accurate ways in computing motion vectors. SUMMARY This disclosure describes motion estimation techniques that can improve video encoding. In particular, this disclosure proposes a non-conventional ways to process video blocks in a frame. Techniques are described below for improving motion estimation. In one embodiment, a motion estimator is described that uses a two dimensional pipeline to generate accurate motion estimation parameters for a current video block. The two dimensional pipeline uses previously calculated motion estimation parameters of relevant neighboring video blocks, including a preceding video block on a same row as the current video block, prior to the generation of the accurate motion estimation parameters. The motion estimation parameters are, for example, motion vectors, motion vector predictors and mode decision. The disclosed motion estimation techniques can improve the encoding by pipelining an engine/module to fetch pixels, an engine/module that performs an integer- WO 2007/038357 PCT/US2006/037130 pixel search, and an engine/module that performs a finer fractional and spatial search over at least two video block rows in a ping pong fashion. During the ping portion, two video blocks from a first block row are processed, while simultaneous processing on another video block row is taking place. During the pong portion, two video blocks from a second block row are processed, while on a first block row another video block is being processed. Processing in this ping-pong fashion allows the integer search engine/module to compute and output more accurate motion vector related cost due to the exact motion vector predictor (MVP) which can only be obtained after fractional and spatial engine/module has finished processing all the neighboring video blocks. The MVP is an early estimate of a desired motion vector and is typically computed based on motion vectors previously calculated for neighboring video blocks. In techniques where only one video block row is processed sequentially the MVP's may be an estimate not an actual value. Using only the estimate limits the accuracy that can be used in motion estimation. One of the advantages of using the disclosed technique is to use all the finer resolution actual values in calculating motion estimation. Another advantage, is that the use of a two dimensional pipeline resolves the problem of needing to wait for one entire row to be processed before processing parts of another row. As such, processing in this ping-pong fashion with a two dimensional pipeline reduces the bandwidth on a communication bus. The number of refreshes to an external memory to update a search area may also significantly be reduced. [0011] The embodiments in this disclosure propose the computation of distortion measures. The embodiments describes a method comprising computing a motion vector predictor based on motion vectors previously calculated for video blocks in proximity to a current video block to be encoded, and using the motion vector predictor in searching for a prediction video block used to encode the current video block. The embodiments further describe a method which uses all actual motion vector predictors to minimize the computation of distortion measures. [0012] These and other techniques described herein may be implemented in a digital video device in hardware, software, firmware, or any combination thereof. If implemented in software, the techniques may be directed to a computer readable medium comprising program code, that when executed, performs one or more of the encoding techniques described herein. Additional details of various embodiments are set forth in the accompanying drawings and the description below. Other features, WO 2007/038357 PCT/US2006/037130 objects and advantages will become apparent from the description and drawings, and from the claims. BRIEF DESCRIPTION OF DRAWINGS [0013] FIG. 1A is a block diagram illustrating an example system in which a source digital video device transmits an encoded bitstream to receive a digital video device. [0014] FIG. IB is a block diagram of two CODEC'S that may be used as described in an embodiment herein. [0015] FIG. 2 is an exemplary block diagram of a basic digital video encoder that may be used in the digital video device illustrated in FIG. 1A or FIG. IB. [0016] FIG. 3 is an exemplary illustration of two partial rows of an image. [0017] FIG. 4 is an exemplary flow chart illustrating the process by which motion estimation and motion vector prediction may be done over two video block rows..... [0018] FIG. 5 is an exemplary illustration of an ME pipeline applied over a partial image row. [0019] FIG. 6 is an exemplary illustration of an ME pipeline applied over two partial image rows. DETAILED DESCRIPTION [0020] This disclosure describes multi-dimensional techniques to calculate both motion vectors (MV's) and motion vector predictors (MVP) which aid in motion estimation techniques that can be used to improve video encoding. Although the techniques are generally described in the context of an overall process for motion estimation, it is understood that one or more of the techniques may be used individually in various scenarios. The MVP is typically computed based on motion vectors previously calculated for neighboring video blocks, e.g., as a median of motion vectors of adjacent video blocks that have been recorded. However, other mathematical functions could alternatively be used to compute the MVP, such as the average of motion vectors or neighboring video blocks or possibly a more complex mathematical function. [0021] FIG. 1A is a block diagram illustrating an example system 100 in which a source device 112a transmits an encoded bitstream via communication link 113 to receive device 114a. Source device 112a and receive device 114a may both be digital video devices. In particular, source device 112a encodes video data consistent with a WO 2007/038357 PCT/US2006/037130 video standard such as the MPEG-4 standard, the ITU H.263 standard, the ITU H.264 standard, or any of a wide variety of other standards that male use of motion estimation in the video encoding. One or both of devices 112a, 114a of system 100 implement motion estimation and compensation techniques, as described in greater detail below, in order to improve the video encoding process. [0022] Communication link 113 may comprise a wireless link, a physical transmission line, fiber optics, a packet based network such as a local area network, wide-area network, or global network such as the Internet, a public switched telephone network (PSTN), or any other communication link capable of transferring data. The communication link 113 may be coupled to a storage media such as CD, DVD etc. Thus, communication link 113 represents any suitable communication medium, or possibly a collection of different networks and links, for transmitting video data from source device 112a to receive device 114a. [0023]..Source device 112a may include a video capture device 1157 such as a video camera, to capture video sequences and store the captured sequences in memory 116. The video sequences may be viewed on display 117. Li particular, video capture device 115 may include a charge coupled device (CCD), a charge injection device, an array of photodiodes, a complementary metal oxide semiconductor (CMOS) device, or any other photosensitive device capable of capturing video images or digital video sequences. [0024] As further examples, video capture device 115 may be a video converter that converts analog video data to digital video data, e.g., from a television, video cassette recorder, camcorder, or another video device. In some embodiments, source device 112a may be configured to transmit real-time video sequences over communication link 115. In that case, receive device 114a may receive the real-time video sequences and display the video sequences to a user. Alternatively, source device 112a may capture and encode video sequences that are sent to receive device 114a as video data files, i.e., not in real-time. Thus, source device 112a and receive device 114a may support applications such as video clip playback, video mail, or video conferencing, e.g., in a mobile wireless network. Devices 112a and 114a may include various other elements that are not specifically illustrated in FIG.. 1A. [0025] Furthermore, source device 112a may be any digital video device capable of encoding and transmitting video data. Source device 112a may also a video encoder 118 to encode the sequences, and a transmitter 120 to transmit the encoded bitstream WO 2007/038357 PCT/US2006/037130 over communication link 115 to source device 114a. Video encoder 118 may include, for example, various hardware, software or firmware, or one or more digital signal processors (DSP) that execute programmable software modules to control the video encoding techniques, as described herein. Associated memory and logic circuitry may be provided to support the DSP in controlling the video encoding techniques. As will be described, video encoder 118 may perform better if accurate values of motion vector predictor (MVPs) are used. [0026] Receive device 114a may take the form of any digital video device capable of receiving and decoding video data. For example, receive device 114a may include a receiver 122 to receive encoded digital video sequences from transmitter 120, e.g., via intermediate links, routers, other network equipment, and like. Receive device 114a also may include a video decoder 124 for decoding the bitstream, and a display device 126 to display the sequences of the decoded bitstream to a user. In some embodiments, however, receive device 114a may not include an integrated display device 114a. In such cases, receive device 114a may serve as a receiver that decodes the received video data to drive a discrete display device, e.g., a television or monitor. [0027] Example devices for source device 112a and receive device 114a include servers located on a computer network, workstations or other desktop computing devices, and mobile computing devices such as laptop computers or personal digital assistants (PDAs). Other examples include digital television broadcasting satellites and receiving devices such as digital televisions, digital cameras, digital video cameras or other digital recording devices, digital video telephones such as mobile telephones having video capabilities, direct two-way communication devices with video capabilities other wireless video devices, and the like. [0028] In some cases, a source device 112b and receive device 114b may each include an encoder/decoder (CODEC) as shown in FIG. IB, for encoding and decoding digital video data. In particular, both source device 112a and receive device 114a may include transmitters and receivers as well as memory and displays. Many of the encoding techniques outlined below are described in the context of a digital video device that includes an encoder. It is understood, however, that the encoder may form part of a CODEC. In that case, the CODEC may be implemented within hardware, software, firmware, a DSP, a microprocessor, an application specific integrated circuit (ASIC), a WO 2007/038357 PCT/US2006/037130 field programmable gate array (FPGA), discrete hardware components, or various combinations thereof. [0029] Video encoder 118 within source device 112a or source device 112b operates on blocks of pixels within a sequence of video frames in order to encode the video data. For example, video encoder 118 may execute motion estimation and motion compensation techniques in which a video frame to be transmitted is divided into blocks of pixels (referred to as video blocks). The video blocks, for purposes of illustration, may comprise any size of blocks, and may vary within a given video sequence. As an example, the ITU H.264 standard supports 16 by 16 video blocks, 16 by 8 video blocks, 8 by 16 video blocks, 8 by 8 video blocks, 8 by 4 video blocks, 4 by 8 video blocks and 4 by 4 video blocks. The use of smaller video blocks in the video encoding can produce better compression in the encoding, and may be specifically used for locations of video frame that include higher levels of detail. [0030] Each pixel in a video block maybe represented by an n-bit value, e.g., 8 bits, which defines visual characteristics of the pixel such as the color and intensity in values of chrominance and luminance. However, motion estimation is often performed only on the luminance component because human vision is more sensitive to changes in luminance than chromaticity. Accordingly, for purposes of motion estimation, the entire n-bit value may quantify luminance for a given pixel. The principles of this disclosure, however, are not limited to the format of the pixels, and may be extended for use with simpler fewer-bit pixel formats or more complex larger-bit pixel formats. [0031] For each video block in the video frame, video encoder 118 of source device 112a or 112b performs motion estimation by searching video blocks stored in memory 116 for one or more preceding video frames already transmitted (or a subsequent video frames) to identify a similar video block, referred to as a prediction video block. In some cases, the prediction video block may comprise the "best prediction block" from the preceding or subsequent video frame, although this disclosure is not limited in that respect. Video encoder 118 performs motion compensation to create a difference block indicative of the differences between the current video block to be encoded and the best prediction block. Motion compensation usually refers to the act of fetching the best prediction block using a motion vector, and then subtracting the best prediction block from an input block to generate a difference block. WO 2007/038357 PCT/US2006/037130 [0032] After the motion compensation process has created the difference block, a series of additional encoding steps are typically performed to encode the difference block. These additional encoding steps may depend on the encoding standard being used. [0033] Once encoded, the encoded difference block can be transmitted along with a motion vector or encoded motion vector that identifies the video block from the previous frame (or subsequent frame) that was used for encoding. In this manner, instead of encoding each frame as an independent picture, video encoder 118 encodes the difference between adjacent frames. Such techniques can significantly reduce the amount of data that needed to accurately represent each frame of a video sequence. [0034] The motion vector may define a pixel location relative to the upper-left-hand corner of the video block being encoded, although other formats for motion vectors could be used. By encoding video blocks using motion vectors, the required bandwidth for transmission of streams of video data can be significantly reduced. [0035] In some cases, video encoder 118 can support intra-frame encoding, in addition to inter-frame encoding. Intra-frame encoding utilizes similarities within frames, referred to as spatial or intra-frame correlation, to further compress the video frames. Intra-frame compression is typically based upon texture encoding for compressing still images, such as discrete cosine transform (DCT) encoding. Intra-frame compression is often used in conjunction with inter-frame compression, but may also be used as an alterative in some implementations. [0036] Receiver 122 of receive device 114a may receive the encoded video data in the form of motion vectors and encoded difference blocks indicative of encoded differences between the video block being encoded and the best prediction block used in motion estimation. In some cases, however, rather than sending motion vectors the difference between the motion vectors and the MVP are transmitted. In any case, decoder 124 can perform video decoding in order to generate video sequences for display to a user via display device 126. The decoder 124 of receive device 114a may also be implemented as an encoder/decoder (CODEC) as shown in FIG IB. In that case, both source device 112b and receive device 114b may be capable of encoding, transmitting, receiving and decoding digital video sequences. [0037] FIG. 2 illustrates an exemplary video encoder that may be used in a device of FIG. 1A or FIG. IB. Frames or part of frames from a video sequence may be placed in an input frame buffer 242 inside a video encoder 118 that may be part of CODEC 24. WO 2007/038357 PCT/US2006/037130 An input frame from input frame buffer 242 may be parsed into blocks (the video blocks may be of any size, but standard square video block sizes are 4x4, 8x8, or 16x16) and sent to video block buffer 243. Video block buffer 243 typically sends a video block to subtractor 244. Subtractor 244 subtracts video block x from the output of switch 246. Switch 246 may switch between intra-coding and inter-coding prediction modes of encoding. If switch 246 is enabling an inter-coding prediction mode, then the resulting difference from x and a video block from a different (previous or subsequent) frame is compressed through texture encoder 247. If switch 246 enables an intra-coding prediction mode, then the resulting difference from x and predicted value from a previous video block in the same frame is compressed through texture encoder 247. [0038] Texture encoder 247 has a DCT block 248 which transforms the input x (the video block or difference block) from the pixel domain to a spatial frequency domain. In the spatial frequency domain, data is represented by DCT block coefficients. The .DCT block coefficients represent-the number and degree of the spatial frequencies detected in the video block. After a DCT is computed, the DCT block coefficients may be quantized by quantizer 250, in a process is known as "block quantization." Quantization of the DCT block coefficients (coming from either the video block or difference video block) removes part of the spatial redundancy from the block. During this "block quantization" process, further spatial redundancy may sometimes be removed by comparing the quantized DCT block coefficients to a threshold. This comparison may take place inside quantizer 250 or another comparator block (not shown). If the magnitude of a quantized DCT block coefficient is less than the threshold, the coefficient is discarded or set to a zero value. [0039] After block quantization, the resulting output may be sent to two separate structures: (1) a texture decoder 265, and (2) an entropy encoder 255. Texture decoder 265 comprises a de-quantizer 266 which aids in the production of a reconstructed video block or frame; to be used with a coding prediction mode. The entropy encoder 255 produces a bitstream for transmission or storage. Entropy encoder 255 may contain a scanner 256 which receives the block quantized output and re-order it for more efficient encoding by variable length coder (VLC) 258. VLC 258 may employ the use of run-length and huffman coding techniques to produce an encoded bit-stream. The encoded bitstream is sent to output buffer 260. The bitstream may be sent to rate controller 262. While maintaining a base quality, rate controller 262 budgets the number of WO 2007/038357 PCT/US2006/037130 quantization bits used by quantizer 250. Entropy encoding is considered a non-lossy form of compression. Non-lossy compression signifies that the data being encoded may be identically recovered if it is decoded by an entropy decoder without the encoded data having been corrupted. Entropy encoder 255 performs non-lossy compression. [0040] Lossy compression means that as a result of the encoding, an input, x, will not produce an identical copy of x even though the encoded input has not been corrupted. The reconstructed input has "lost" part of its information. Texture encoder 247 performs lossy compression. A typical video encoder 118 usually has a local texture decoder 265 to aid in the compensation of both the inter-coding and intra-coding prediction modes. De-quantizer 266, inverse DCT 268, and the output of switch 246 that is sent to adder 269 work together to decode the output of texture encoder 247 and reconstruct the input x that went into texture encoder 247. The reconstructed input, y, looks similar to x but is not exactly x. A general video "decoder" typically comprises the functionality of the de-quantizer 266* inverse DGT 687 and the output of switch 246 that is sent to adder 269. [0041] The reconstructed input may be sent to memory buffer 281. Inside memory buffer 281 there may be two memory buffers: (1) reconstructed new frame buffer 282; and (2) reconstructed old frame buffer 284. Reconstructed new frame buffer 282, stores the currently processed reconstructed frame (or partial frame). Reconstructed old frame buffer 284 stores a past processed reconstructed frame. The past processed reconstructed frame is used as a (reconstructed) reference frame. The reconstructed reference frame may be a frame that is before or after the current frame in input frame buffer 242. The current frame (or a video block from the current frame) or differences between the current frame and the reconstructed reference frame (or a video block from the difference block) is what is "currently" being encoded. After the current frame has finished encoding and before the next frame in input from input frame buffer 242 is fetched to be encoded, the reconstructed old frame buffer 284 is updated with a copy with the contents of the reconstructed new frame buffer 282. [0042] Reconstructed new frame buffer 282 may send the reconstructed video block it received to be used in spatial predictor 286. Reconstructed old frame buffer 284 sends a past processed reconstructed video block to MEC (motion estimation and compensation block) 287. MEC block comprises motion estimator 88 and motion compensator 290. Motion estimator 288 generates motion vectors (MV) 92 and motion vector predictors WO 2007/038357 PCT/US2006/037130 (MVP) 294 that may be used by motion compensator 290 to compensate for differences from other frames than the one being encoded. MVs 292 may also be used by entropy encoder 255. In some standards, such as ITU H.264, the output of spatial predictor 286 is used in intra-frame prediction mode and fed back both to subtracter 244 and adder 269. In some standards, such as MPEG-4 or JPEG, there is no spatial predictor 286. [0043] FIG. 3 illustrates two partial block rows of any image or frame. As an example, let block row N-l and N be block row 3 and block row 4. In block row 3 330 there are nine video blocks. For illustrative purposes 16x16 blocks may be used throughout the description of the disclosure. Hence, macroblocks (MB's), 331-339 are in row 3 330, and in row 4 340 there are nine MB's 341-749. The MB's are drawn showing both the block row number and the location relative to the Mth macroblock. M denotes a current macroblock. Typically block row 3 is processed before block row 4. Within this disclosure processing a block row means processing a row of macroblocks as outlined in row 3 330 and row 4340. In general, it can also mean the processing Of any size of video block row. [0044] When calculating motion vectors relative to macroblock (MB) 345 in various standards, such as H.264, MPEG-4, and H.263, etc., it may be desired to have prior knowledge of MB modes and motion vectors of the neighboring MB 344, MB 335, and MB 336 (or MB 334 when MB 336 is not available). For example, in H.264, the INTER modes for a P-frame can be INTER 16x16, INTER 16x8, INTER 8x16, INTER 8x8. If it is an INTER 8x8 mode then a further partition can be made to select INTER 8x4, INTER 4x8 or INTER 4x4 modes. The mode is not only dependent on they type (INTER) but also the size. There can also be INTRA mode and SKIP mode. Although there are other conditions that invoke SKIP mode, one condition that may invoke SKIP mode is when the MV equals the MVP. [0045] Motion estimation, for example, typically requires a larger amount of computational resources than any other process of video encoding. For this reason, it is highly desirable to perform motion estimation in a manner that can reduce computational complexity and also help in improving the compression ratio. The motion estimation techniques described herein may advance these goals by using a search scheme that performs the searching at multiple spatial resolutions, thereby reducing the computational complexity without any loss in accuracy. In addition, a cost function, also known as a distortion measure is proposed, that includes the cost of WO 2007/038357 PCT/US2006/037130 encoding motion vectors. A motion estimator may also use multiple candidate locations of a search space to improve the accuracy of video encoding, and the search area around the multiple candidates may be programmable, thus making the process scalable with fame rate and picture sizes. Finally, a motion estimator may also combine cost functions for many small square video blocks, e.g., 4 by 4 video blocks, to obtain the cost for the various larger block shapes, e.g., 4 by 8 video blocks, 8 by 4 video blocks, 8 by 8 video blocks, 8 by 16 video blocks, 16 by 8 video blocks, 16 by 16 video blocks, and so forth.. For many operations and computations, a motion vector predictor (MVP) is used to add a cost-factor for motion vectors deviating from the motion vector predictor. The MVP may also provide an additional initial motion vector, which can be used to define searches, particularly at high resolution stages of a multi-stage search. The calculation of distortion measure values that depend at least in part the motion vector prediction values is part of the cost-factor. Distortion measure values may also help quantify the number of bits needed to encode different motion vectors. [0046] FIG. 4 is an exemplary flow chart describing the process of how motion estimation can be done throughout a frame. First a setup procedure 400 is initiated. As part of the setup a reference (a past or future) frame is loaded 402 into a memory. For example, the memory may be a local memory or off-memory candidate RAM. Then a current frame is loaded 404 into a memory. For example, the memory may also be part of the local memory or off-memory encode RAM. This is followed by a selection of the current macroblock M in a block row. Then a search space is identified 406 around the current macroblock M. Once a search space has been identified from a reference frame 408, the act of processing a 2-D stage pipeline over two video block rows in a current frame 410 takes place. This is an improvement over the current art which only considers processing of one video block row at a time via a 1-D stage pipeline. After each 2-D stage pipeline act is performed, decision block 412 checks if the end of two block rows has been reached. If the end of the two block rows has not been reached (NO), then the search space in the reference frame is refreshed 414, and processing of the 2-D stage pipeline over two block rows in a current frame 410 continues. If the end of two block rows has been reached (YES) then another check is performed by decision block 416. Decision block 416 checks if it is the end of the last two block rows in the WO 2007/038357 PCT/US2006/037130 current frame. If it is not the end of the last two block rows in a current frame (NO), then the act of incrementing to the next two block rows 418 is performed, and processing of the 2-D stage pipeline over two block rows in a current frame 410 continues. If is the end of the last two block rows in a current frame (YES) processing of the current frame ends. [0047] Motion estimation (ME) involves processing a video block by at least two engines. The first engine is a fetch engine (FE), and may fetch video blocks from a memory to be encoded. The second engine uses a distortion metric to find a similar matching block which minimizes the distortion. The second engine is an integer search engine (ISE), and may employ a hierarchical search with search being performed in a finer resolution as it proceeds. There may also be a third engine, a fractional and spatial search engine (FSE), which may use a finer resolution search to find a similar matching block which minimizes the distortion. The FSE may or may not use the results of the second engine, as a starting point. For exemplary purposes,- description of the motion estimation technique is done with three engines, the fetch Engine (FE), the integer Search Engine (ISE), and the fractional and spatial search engine (FSE). These three engines process a video block in a sequential fashion. Typically, to minimize the delay of the processing, the three engines are pipelined over three stages. That is to say, an FE, ISE, and FSE all operate in parallel in a three stage pipeline. For example, during stage 1, an FE fetches from a memory the current macroblock M, while simultaneously, an ISE positioned with an integer pixel anchor point at the corner of a macroblock M-l tries to find the best match macroblock from a reference (a past or future) frame. Still during stage 1, simultaneously operating at the same time as the FE and ISE engines, an FSE positioned at a fractional pixel anchor point in macroblock M-2 tries to find the best match macroblock from a reference (a past or future) frame. It takes three stages to complete the complete processing of one video block to generate a motion vector. At the end of three stages there will have been three fetches, three integer searches performed and three fractional and spatial searches performed. Typically, the one dimensional (ID) three stage pipeline operates sequentially over one video block row. No processing takes place in the second block row until the whole video block row has been completely processed. [0048] FIG 5 illustrates a conceptual illustration of the ID three stage pipeline. As can be seen most of row 3 330 is being processed. In the upper left comer of FIG 5, two of WO 2007/038357 PCT/US2006/037130 the three engines are running during stage 1 501 of the 3, M-3 MB pipeline 520. FE operates on macroblock 3, M-3 332, and ISE operates on macroblock 3, M-4 331. Stage 2 502 of the 3, M-3 MB pipeline 520 and stage 3 503 of the 3, M-3 MB pipeline 520 complete the generation of motion vectors after the FSE is completed on the 3, M-3 MB (macroblock) 332. As mentioned previously, in general, at each stage a fetch, an integer search, and a fractional and spatial search is performed. With the exception of the initial fetch of the 3, M-4 macroblock 331, all macroblocks (MBs) in row 3 330 of FIG. 3 are processed in FIG. 5. For 1-D pipelines 521 - 527, stagel of the next pipeline is always stage 2 of the previous pipeline. Stage 2 of the next pipeline is stage 3 of the previous pipeline. Thus, stages 502-510 may be a stage 1, stage 2 or stage 3 stage in a pipeline, depending on which pipeline is currently operating. In general it takes N+2 stages to process N macroblocks (i.e. video blocks) using a ID pipeline. [0049] For a given macroblock, the ID pipeline technique constrains the current ISE to not use the FSE result.of the previous motion block to its advantage; For example," when the FE is fetching the encode data and updating the memory for MB 335, ISE is working on integer search for MB 334 and FSE is performing fractional search and spatial estimation for MB 333. An inherent problem with the ID pipeline is that ISE does not know the modes and final motion vectors (MV's) of its left neighboring MB and thus is not able to obtain an accurate estimate of the motion vector predictors (MVP's). As a result the motion vector computation may be slightly off. Hence, by using a ID pipeline technique, there exists a mutual dependency between motion estimation (generating MV's) and mode decision (generating MVP's). [0050] A motion estimation technique that resolves the mutual dependencies between motion estimation and mode decision is illustrated in FIG 6. FIG. 6 is an exemplary diagram illustrating a three stage two-dimensional (2D) pipeline, motion estimation technique that allows the ISE prior knowledge of all the relevant neighboring MB modes and MV's. In FIG. 6, the two video block rows are processed in a ping pong fashion. During the ping portion, two video blocks from a first video block row are processed, while simultaneous processing on another video block row is taking place. During the pong portion, two video blocks from a second video block row are processed, while on a first video block row another video block is being processed. [0051] In the upper left corner of FIG. 6, illustrates stage 1 601 of the 3, M-0 MB 2D pipeline 621. FE operates on macroblock 3, M-0 335, and ISE operates on macroblock WO 2007/038357 PCT/US2006/037130 4, M-3 342, while FSE operates on macroblock 3, M-l 334. During stage 2 602 of the 3, M-0 MB 2D pipeline 621, FE operatres on macroblock 4,M-2 343, and ISE operates on macroblock 3, M-0 335, while FSE operates on macroblock 4, M-3 342. During stage 3 603 of the 3, M-0 MB 2D pipeline 621, FE operatres on macroblock 3.M+1 336, and ISE operates on macroblock 4, M-2 343, while FSE operates on macroblock 3, M-0 335. After completion of the 3, M-0 MB 2D pipeline 621 generation of motion vectors after the FSE is completed on the 3, M-0 MB 335. Stages 602-609 may be used as described above to complete 2D pipelines 622-627. Every "ping/pong", see FIG. 6, represents an FE/FSE operation on one row while there is an ISE on another row. The ping pong terminology is associated with the switching of roles between rows. For example, during the "ping" FE/FSE operate on row 3, and ISE operates on row 4. During the "pong" FE/FSE operate on row 4, and ISE operates on row 3. [0052] An advantage of using the 2D pipeline is that two macroblocks from at least two different rows but close in proximity (in this example, horizontal position) may share a majority of the search area. Since the pong can re-use search area from the ping there will be less fetches required. Processing in this ping-pong fashion reduces the bandwidth on a communication bus. The number of refreshes to an external memory to update a search area may also significantly be reduced. The number of refreshes to an external memory to update a search area may also significantly be reduced. [0053] In addition, macroblock 345 can use the mode decision from relevant neighboring macroblocks, 344, 335, and 334 (or 336 if 334 is not available). The use of mode decision means that the motion vector predictor from each of those relevant neighboring macroblocks may aid in producing an accurate estimate of the motion vector. Without the 2D ping pong pipeline, an accurate motion vector predictor from neighboring macroblock 344 may not be available in the computation of macroblock's 345 motion vector generation. Hence, the two-dimensional pipeline resolves a mutual dependency between a motion estimation video block and a mode decision. [0054] A number of different embodiments have been described. The techniques may be capable of improving video encoding by improving motion estimation. The techniques may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the techniques may be directed to a computer readable medium comprising program code, that when executed in a device that encodes video sequences, performs one or more of the methods mentioned above. In that case, WO 2007/038357 PCT/US2006/037130 the computer readable medium may comprise random access memory (RAM) such as synchronous dynamic random access memory (SDRAM), read-only memory (ROM), non-volatile random access memory (NVRAM), electrically erasable programmable read-only memory (EEPROM), FLASH memory, and the like. [0055] The program code may be stored on memory in the form of computer readable instructions. In that case, a processor such as a DSP may execute instructions stored in memory in order to carry out one or more of the techniques described herein. In some cases, the techniques may be executed by a DSP that invokes various hardware components such as a motion estimator to accelerate the encoding process. In other cases, the video encoder may be implemented as a microprocessor, one or more application specific integrated circuits (ASICs), one or more field programmable gate arrays (FPGAs), or some other hardware-software combination. These and other embodiments are within the scope of the following claims. WO 2007/038357 PCT/US2006/037130 We claim: 1. A device having a motion estimator using a two dimensional pipeline to generate accurate motion vectors, motion vector predictors, and mode decision for a current video block, the two dimensional pipeline using completed motion vector, motion vector predictors, and mode decision calculations of relevant neighboring video blocks, including a preceding video block on a same video block row as the current video block, prior to the generation of the accurate motion vector, motion vector predictor, and mode decision of the current video block. 2. The device of claim 1, wherein the two dimensional pipeline comprises using a fetch engine and a fractional spatial search engine operating on a video block row, while an integer search engine operates on another video block row. 3. The device of claim 1, wherein the two dimensional pipeline involves processing at least two stages in the pipeline. 4. The device of claim 1, wherein the video block is a macroblock, an 8x8 block, an 8x4 block, a 4x8 block, a 4x4 block, or a 2x2 block. 5. In a device including a motion estimator, a method of performing two dimensional pipeline processing to generate accurate motion estimation parameters for a current video block, the two dimensional pipeline processing includes identifying previously calculated motion estimation parameters of neighboring video blocks, including a preceding video block on a same video block row as the current video block, prior to the generation of the accurate motion estimation parameters of the current video block. 6. The method of claim 5, wherein the motion estimation parameters are a motion vector from the current video block, motion vector predictors associated with the current video block, and mode decision.. WO 2007/038357 PCT/US2006/037130 7. The method of claim 5, wherein the two dimensional pipeline processing comprises using a fetch engine and a fractional spatial search engine operating on a video block row, while an integer search engine operates on another video block row. 8. The method of claim 5, wherein the two dimensional pipeline processing involves at least two stages in the pipeline. 9. The method of claim 5, wherein the video block is a macroblock, an 8x8 block, an 8x4 block, a 4x8 block, a 4x4 block, or a 2x2 block. 10. A device having a motion estimator including means for performing two dimensional pipeline processing to generate accurate motion estimation parameters for a current video block, the two dimensional pipeline processing having means for identifying previously, calculated motion estimation parameters of relevant neighboring video blocks, including a preceding video block on a same video block row as the current video block, prior to the generation of the accurate motion estimation parameters of the current video block. 11. The device of claim 10, wherein the motion estimation parameters are a motion vector from the current video block, motion vector predictors associated with the current video block, and mode decision. 12. The device of claim 10, wherein the two dimensional pipeline processing comprises using a fetch engine and a fractional spatial search engine operating on a video block row, while an integer search engine operates on another video block row. 13. The device of claim 10, wherein the two dimensional pipeline processing involves at least two stages in the pipeline. 14. The device of claim 10, wherein the video block is a macroblock, an 8x8 block, an 8x4 block, a 4x8 block, a 4x4 block, or a 2x2 block. WO 2007/038357 PCT/US2006/037130 15. A computer program product comprising a computer readable medium having instructions stored thereon when executed, generate a motion vector for a current video block, the computer program product comprising sets of instructions for: performing two dimensional pipeline processing to generate accurate motion estimation parameters for a current video block, the two dimensional pipeline processing further includes instructions for identifying previously calculated motion estimation parameters of neighboring video blocks, including a preceding video block on a same row as the current video block, prior to the generation of the accurate motion estimation parameters of the current video block. 16. The computer program product of claim 15, wherein the motion estimation parameters are a motion vector from the current video block, motion vector predictors associated with the current video block, and mode decision. 17. The computer program product of claim 15, wherein the two dimensional pipeline processing comprises using a fetch engine and a fractional spatial search engine operating on a video block row, while an integer search engine operates on another video block row. 18. The computer program product of claim 15, wherein the two dimensional pipeline processing involves at least two stages in the pipeline. 19. The computer program product of claim 15, wherein the video block is a macroblock, an 8x8 block, an 8x4 block, a 4x8 block, a 4x4 block, or a 2x2 block. 20. A method of video encoding comprising: computing a motion vector predictor based on motion vectors previously calculated for video blocks in proximity to a current video block to be encoded; using the motion vector predictor in searching for a prediction block used to encode the current video block; and calculation of at least one motion vector predictor and at least one motion vector by processing at least two video block rows using a two dimensional pipeline. WO 2007/038357 PCT/US2006/037130 21. The method of claim 20, further comprising generating a difference block indicative of differences between the current video block to be encoded and the best prediction block. 22. The method of claim 21, further comprising identifying a motion vector to the prediction video block used to encode the current video block including calculating distortion measure values that depend at least in part the motion vector prediction value. 23. The method of claim 22, wherein the distortion measure values quantify number of bits needed to encode different motion vectors. 24. The method of claim 20, further comprising performing searches in stages at different spatial resolutions to identify the-motion vector to the best prediction block used to encode the current video block. 25. The device of claim 20, wherein the two dimensional pipeline comprises using a fetch engine and a fractional spatial search engine operating on a video block row, while an integer search engine operates on another vide block row. 26. A video encoding device comprising: a motion estimator that identifies a motion vector to a best prediction block used to encode a current video block, including calculating distortion measure values that depend at least in part on an amount of data associated with different motion vectors; a motion compensator that generates a difference block indicative of differences between the current video block to be encoded and the prediction video block; and calculation of at least one motion vector predictor and at least one motion vector by using a two dimensional pipeline over at least two video block rows. WO 2007/038357 PCT/US2006/037130 27. The video encoding device of claim 26, wherein the motion estimator performs searches in stages at different spatial resolutions to identify the motion vector to the best prediction block used to encode the current video block. 28. The video encoding device of claim 26, wherein the video encoding device computes a motion vector predictor based on motion vectors previously calculated for video blocks in proximity to the current video block to be encoded, wherein the motion vector predictor value defines a search in at least one of the stages and is also used to calculate the distortion measure values. ABSTRACT "MULTI-DIMENSIONAL NEIGHBORING BLOCK PREDICTION FOR VIDEO ENCODING" The motion estimation techniques and video encoding device(s) described use a two dimensional pipeline to generate accurate motion estimation parameters for a current video block. The two dimensional pipeline uses previously calculated motion estimation parameters of relevant neighboring video blocks, including a preceding video block on a same row as the current video block, prior to the generation of the accurate motion vectors, motion vector predictors, and mode decision of the current video block The use of the two dimensional pipeline allows accurate motion vector prediction from neighboring video blocks previously not available, in the computation of motion vectors. Three engines may be used in the two dimensional pipeline, a fetch engine, an integer search engine and a fractional and spatial search engine. While the fetch engine and fraction and spatial search engine operate on one row, the integer search engine operates on another row.

Full Text

FORM 2
THE PATENTS ACT, 1970
(39 of 1970)
&
THE PATENTS RULES, 2003
COMPLETE SPECIFICATION
(See section 10, rule 13)
"MULTI-DIMENSIONAL NEIGHBORING BLOCK PREDICTION FOR VIDEO
ENCODING"
QUALCOMM INCORPORATED, a company incorporated in the state of Delaware, U.S., of 5775 Morehouse Drive, San Diego, California 92121-1714, U.S.A
The following specification particularly describes the invention and the manner in which it is to be performed.

WO 2007/038357 PCT/US2006/037130
MULTI-DIMENSIONAL NEIGHBORING BLOCK PREDICTION FOR VIDEO
ENCODING
This present application claims the benefit of U.S. Provisional Application No. 60/719, 891, filed on September 22,2005, and assigned to the assignee hereof, which is hereby incorporated herein by reference in its entirety.
TECHNICAL FIELD
[0001] What is described herein relates to digital video processing and, more particularly, encoding of video sequences.
BACKGROUND
[0002] Digital video capabilities can be incorporated into a wide range of devices, including digital televisions", digital direct broadcast systems, wireless communication devices, personal digital assistants (PDAs), laptop computers, desktop computers, digital cameras, digital recording devices, cellular or satellite radio telephones, and the like. Digital video devices can provide significant improvements over conventional analog video systems in creating, modifying, transmitting, storing, recording and playing full motion video sequences.
[0003] A number of different video encoding standards have been established for encoding digital video sequences. The Moving Picture Experts Group (MPEG), for example, has developed a number of standards including MPEG-1, MPEG-2 and MPEG-4. Other standards include the International Telecommunication Union (ITU) H.263 standard, QuickTime™ technology developed by Apple Computer of Cupertino California, Video for Windows™ developed by Microsoft Corporation of Redmond, Washington, Indeo™ developed by Intel Corporation, RealVideo™ from RealNetworks, Inc. of Seattle, Washington, and Cinepak™ developed by SuperMac, Inc. New standards continue to emerge and evolve, including the ITU H.264 standard and a number of proprietary standards.
[0004] Many video encoding standards allow for improved transmission rates of video sequences by encoding data in a compressed fashion. Compression can reduce the overall amount of data that needs to be transmitted for effective transmission of video

WO 2007/038357 PCT/US2006/037130
frames. Most video encoding standards, for example, utilize graphics and video compression techniques designed to facilitate video and image transmission over a narrower bandwidth than can be achieved without the compression.
[0005] The MPEG standards and the ITU H.263 and ITU H.264 standards, for example, support video encoding techniques that utilize similarities between successive video frames, referred to as temporal or inter-frame correlation, to provide inter-frame compression. The inter-frame compression techniques exploit data redundancy across frames by converting pixel-based representations of video frames to motion representations. In addition, some video encoding techniques may utilize similarities within frames, referred to as spatial or intra-frame correlation, to further compress the video frames.
[0006] In order to support compression, a digital video device includes an encoder for compressing digital video sequences, and a decoder for decompressing the digital video sequences. In many cases, the encoder and decoder form an integrated encoder/decoder (CODEC) that operates on blocks of pixels within frames that define the sequence of video images. In the International Telecommunication Union (ITU) H.264 standard, for example, the encoder typically divides a video frame to be transmitted into video blocks referred to as "macroblocks" (MB's) which may comprise 16 by 16 pixel arrays. The ITU H.264 standard supports 16 by 16 video blocks, 16 by 8 video blocks, 8 by 16 video blocks, 8 by 8 video blocks, 8 by 4 video blocks, 4 by 8 video blocks and 4 by 4 video blocks. Other standards may support differently sized video blocks.
[0007] For each video block in the video frame, an encoder searches similarly sized video blocks of one or more immediately preceding video frames (or subsequent frames) to identify the most similar video block, referred to as the "best prediction block." The process of comparing a current video block to video blocks of other frames is generally referred to as motion estimation. Once a "best prediction block" is identified for a video block, the encoder can encode the differences between the current video block and the best prediction block. This process of encoding the differences between the current video block and the best prediction block includes a process referred to as motion compensation. Motion compensation comprises a process of creating a difference block, indicative of the differences between the current video block to be encoded and the best prediction block. Motion compensation usually refers to the

WO 2007/038357 PCT/US2006/037130
act of fetching the best prediction block using a motion vector, and then subtracting the best prediction block from an input block to generate a difference block. [0008] After motion compensation has created the difference block, a series of additional encoding steps are typically performed to encode the difference block. These additional encoding steps may depend on the encoding standard being used. In MPEG4 compliant encoders, for example, the additional encoding steps may include an 8x8 discrete cosine transform, followed by scalar quantization, followed by a raster-to-zigzag reordering, followed by run-length encoding, followed by Huffman encoding. An encoded difference block can be transmitted along with a motion vector that indicates which video block from the previous frame (or subsequent frame) was used for the encoding. A decoder receives the motion vector and the encoded difference block, and decodes the received information to reconstruct the video sequences. [0009] It is highly desirable to simplify and improve the encoding process. To this end, a wide variety of encoding techniques have been developed. Because motion estimation is one of the most computationally intensive processes in video encoding, improvements to motion estimation can provide notable improvements in the video encoding process.
[0010] It is desirable to find more efficient and accurate ways in computing motion vectors.
SUMMARY
This disclosure describes motion estimation techniques that can improve video encoding. In particular, this disclosure proposes a non-conventional ways to process video blocks in a frame. Techniques are described below for improving motion estimation. In one embodiment, a motion estimator is described that uses a two dimensional pipeline to generate accurate motion estimation parameters for a current video block. The two dimensional pipeline uses previously calculated motion estimation parameters of relevant neighboring video blocks, including a preceding video block on a same row as the current video block, prior to the generation of the accurate motion estimation parameters. The motion estimation parameters are, for example, motion vectors, motion vector predictors and mode decision.
The disclosed motion estimation techniques can improve the encoding by pipelining an engine/module to fetch pixels, an engine/module that performs an integer-

WO 2007/038357 PCT/US2006/037130
pixel search, and an engine/module that performs a finer fractional and spatial search over at least two video block rows in a ping pong fashion. During the ping portion, two video blocks from a first block row are processed, while simultaneous processing on another video block row is taking place. During the pong portion, two video blocks from a second block row are processed, while on a first block row another video block is being processed. Processing in this ping-pong fashion allows the integer search engine/module to compute and output more accurate motion vector related cost due to the exact motion vector predictor (MVP) which can only be obtained after fractional and spatial engine/module has finished processing all the neighboring video blocks. The MVP is an early estimate of a desired motion vector and is typically computed based on motion vectors previously calculated for neighboring video blocks. In techniques where only one video block row is processed sequentially the MVP's may be an estimate not an actual value. Using only the estimate limits the accuracy that can be used in motion estimation. One of the advantages of using the disclosed technique is to use all the finer resolution actual values in calculating motion estimation. Another advantage, is that the use of a two dimensional pipeline resolves the problem of needing to wait for one entire row to be processed before processing parts of another row. As such, processing in this ping-pong fashion with a two dimensional pipeline reduces the bandwidth on a communication bus. The number of refreshes to an external memory to update a search area may also significantly be reduced.
[0011] The embodiments in this disclosure propose the computation of distortion measures. The embodiments describes a method comprising computing a motion vector predictor based on motion vectors previously calculated for video blocks in proximity to a current video block to be encoded, and using the motion vector predictor in searching for a prediction video block used to encode the current video block. The embodiments further describe a method which uses all actual motion vector predictors to minimize the computation of distortion measures.
[0012] These and other techniques described herein may be implemented in a digital video device in hardware, software, firmware, or any combination thereof. If implemented in software, the techniques may be directed to a computer readable medium comprising program code, that when executed, performs one or more of the encoding techniques described herein. Additional details of various embodiments are set forth in the accompanying drawings and the description below. Other features,

WO 2007/038357 PCT/US2006/037130
objects and advantages will become apparent from the description and drawings, and from the claims.
BRIEF DESCRIPTION OF DRAWINGS
[0013] FIG. 1A is a block diagram illustrating an example system in which a source
digital video device transmits an encoded bitstream to receive a digital video device.
[0014] FIG. IB is a block diagram of two CODEC'S that may be used as described in
an embodiment herein.
[0015] FIG. 2 is an exemplary block diagram of a basic digital video encoder that may
be used in the digital video device illustrated in FIG. 1A or FIG. IB.
[0016] FIG. 3 is an exemplary illustration of two partial rows of an image.
[0017] FIG. 4 is an exemplary flow chart illustrating the process by which motion
estimation and motion vector prediction may be done over two video block rows.....
[0018] FIG. 5 is an exemplary illustration of an ME pipeline applied over a partial
image row.
[0019] FIG. 6 is an exemplary illustration of an ME pipeline applied over two partial
image rows.
DETAILED DESCRIPTION
[0020] This disclosure describes multi-dimensional techniques to calculate both motion vectors (MV's) and motion vector predictors (MVP) which aid in motion estimation techniques that can be used to improve video encoding. Although the techniques are generally described in the context of an overall process for motion estimation, it is understood that one or more of the techniques may be used individually in various scenarios. The MVP is typically computed based on motion vectors previously calculated for neighboring video blocks, e.g., as a median of motion vectors of adjacent video blocks that have been recorded. However, other mathematical functions could alternatively be used to compute the MVP, such as the average of motion vectors or neighboring video blocks or possibly a more complex mathematical function.
[0021] FIG. 1A is a block diagram illustrating an example system 100 in which a source device 112a transmits an encoded bitstream via communication link 113 to receive device 114a. Source device 112a and receive device 114a may both be digital video devices. In particular, source device 112a encodes video data consistent with a

WO 2007/038357 PCT/US2006/037130
video standard such as the MPEG-4 standard, the ITU H.263 standard, the ITU H.264 standard, or any of a wide variety of other standards that male use of motion estimation in the video encoding. One or both of devices 112a, 114a of system 100 implement motion estimation and compensation techniques, as described in greater detail below, in order to improve the video encoding process.
[0022] Communication link 113 may comprise a wireless link, a physical transmission line, fiber optics, a packet based network such as a local area network, wide-area network, or global network such as the Internet, a public switched telephone network (PSTN), or any other communication link capable of transferring data. The communication link 113 may be coupled to a storage media such as CD, DVD etc. Thus, communication link 113 represents any suitable communication medium, or possibly a collection of different networks and links, for transmitting video data from source device 112a to receive device 114a.
[0023]..Source device 112a may include a video capture device 1157 such as a video camera, to capture video sequences and store the captured sequences in memory 116. The video sequences may be viewed on display 117. Li particular, video capture device 115 may include a charge coupled device (CCD), a charge injection device, an array of photodiodes, a complementary metal oxide semiconductor (CMOS) device, or any other photosensitive device capable of capturing video images or digital video sequences. [0024] As further examples, video capture device 115 may be a video converter that converts analog video data to digital video data, e.g., from a television, video cassette recorder, camcorder, or another video device. In some embodiments, source device 112a may be configured to transmit real-time video sequences over communication link 115. In that case, receive device 114a may receive the real-time video sequences and display the video sequences to a user. Alternatively, source device 112a may capture and encode video sequences that are sent to receive device 114a as video data files, i.e., not in real-time. Thus, source device 112a and receive device 114a may support applications such as video clip playback, video mail, or video conferencing, e.g., in a mobile wireless network. Devices 112a and 114a may include various other elements that are not specifically illustrated in FIG.. 1A.
[0025] Furthermore, source device 112a may be any digital video device capable of encoding and transmitting video data. Source device 112a may also a video encoder 118 to encode the sequences, and a transmitter 120 to transmit the encoded bitstream

WO 2007/038357 PCT/US2006/037130
over communication link 115 to source device 114a. Video encoder 118 may include, for example, various hardware, software or firmware, or one or more digital signal processors (DSP) that execute programmable software modules to control the video encoding techniques, as described herein. Associated memory and logic circuitry may be provided to support the DSP in controlling the video encoding techniques. As will be described, video encoder 118 may perform better if accurate values of motion vector predictor (MVPs) are used.
[0026] Receive device 114a may take the form of any digital video device capable of receiving and decoding video data. For example, receive device 114a may include a receiver 122 to receive encoded digital video sequences from transmitter 120, e.g., via intermediate links, routers, other network equipment, and like. Receive device 114a also may include a video decoder 124 for decoding the bitstream, and a display device 126 to display the sequences of the decoded bitstream to a user. In some embodiments, however, receive device 114a may not include an integrated display device 114a. In such cases, receive device 114a may serve as a receiver that decodes the received video data to drive a discrete display device, e.g., a television or monitor.
[0027] Example devices for source device 112a and receive device 114a include servers located on a computer network, workstations or other desktop computing devices, and mobile computing devices such as laptop computers or personal digital assistants (PDAs). Other examples include digital television broadcasting satellites and receiving devices such as digital televisions, digital cameras, digital video cameras or other digital recording devices, digital video telephones such as mobile telephones having video capabilities, direct two-way communication devices with video capabilities other wireless video devices, and the like.
[0028] In some cases, a source device 112b and receive device 114b may each include an encoder/decoder (CODEC) as shown in FIG. IB, for encoding and decoding digital video data. In particular, both source device 112a and receive device 114a may include transmitters and receivers as well as memory and displays. Many of the encoding techniques outlined below are described in the context of a digital video device that includes an encoder. It is understood, however, that the encoder may form part of a CODEC. In that case, the CODEC may be implemented within hardware, software, firmware, a DSP, a microprocessor, an application specific integrated circuit (ASIC), a

WO 2007/038357 PCT/US2006/037130

field programmable gate array (FPGA), discrete hardware components, or various combinations thereof.
[0029] Video encoder 118 within source device 112a or source device 112b operates on blocks of pixels within a sequence of video frames in order to encode the video data. For example, video encoder 118 may execute motion estimation and motion compensation techniques in which a video frame to be transmitted is divided into blocks of pixels (referred to as video blocks). The video blocks, for purposes of illustration, may comprise any size of blocks, and may vary within a given video sequence. As an example, the ITU H.264 standard supports 16 by 16 video blocks, 16 by 8 video blocks, 8 by 16 video blocks, 8 by 8 video blocks, 8 by 4 video blocks, 4 by 8 video blocks and 4 by 4 video blocks. The use of smaller video blocks in the video encoding can produce better compression in the encoding, and may be specifically used for locations of video frame that include higher levels of detail.
[0030] Each pixel in a video block maybe represented by an n-bit value, e.g., 8 bits, which defines visual characteristics of the pixel such as the color and intensity in values of chrominance and luminance. However, motion estimation is often performed only on the luminance component because human vision is more sensitive to changes in luminance than chromaticity. Accordingly, for purposes of motion estimation, the entire n-bit value may quantify luminance for a given pixel. The principles of this disclosure, however, are not limited to the format of the pixels, and may be extended for use with simpler fewer-bit pixel formats or more complex larger-bit pixel formats.
[0031] For each video block in the video frame, video encoder 118 of source device 112a or 112b performs motion estimation by searching video blocks stored in memory 116 for one or more preceding video frames already transmitted (or a subsequent video frames) to identify a similar video block, referred to as a prediction video block. In some cases, the prediction video block may comprise the "best prediction block" from the preceding or subsequent video frame, although this disclosure is not limited in that respect. Video encoder 118 performs motion compensation to create a difference block indicative of the differences between the current video block to be encoded and the best prediction block. Motion compensation usually refers to the act of fetching the best prediction block using a motion vector, and then subtracting the best prediction block from an input block to generate a difference block.

WO 2007/038357 PCT/US2006/037130
[0032] After the motion compensation process has created the difference block, a series of additional encoding steps are typically performed to encode the difference block. These additional encoding steps may depend on the encoding standard being used. [0033] Once encoded, the encoded difference block can be transmitted along with a motion vector or encoded motion vector that identifies the video block from the previous frame (or subsequent frame) that was used for encoding. In this manner, instead of encoding each frame as an independent picture, video encoder 118 encodes the difference between adjacent frames. Such techniques can significantly reduce the amount of data that needed to accurately represent each frame of a video sequence. [0034] The motion vector may define a pixel location relative to the upper-left-hand corner of the video block being encoded, although other formats for motion vectors could be used. By encoding video blocks using motion vectors, the required bandwidth for transmission of streams of video data can be significantly reduced. [0035] In some cases, video encoder 118 can support intra-frame encoding, in addition to inter-frame encoding. Intra-frame encoding utilizes similarities within frames, referred to as spatial or intra-frame correlation, to further compress the video frames. Intra-frame compression is typically based upon texture encoding for compressing still images, such as discrete cosine transform (DCT) encoding. Intra-frame compression is often used in conjunction with inter-frame compression, but may also be used as an alterative in some implementations.
[0036] Receiver 122 of receive device 114a may receive the encoded video data in the form of motion vectors and encoded difference blocks indicative of encoded differences between the video block being encoded and the best prediction block used in motion estimation. In some cases, however, rather than sending motion vectors the difference between the motion vectors and the MVP are transmitted. In any case, decoder 124 can perform video decoding in order to generate video sequences for display to a user via display device 126. The decoder 124 of receive device 114a may also be implemented as an encoder/decoder (CODEC) as shown in FIG IB. In that case, both source device 112b and receive device 114b may be capable of encoding, transmitting, receiving and decoding digital video sequences.
[0037] FIG. 2 illustrates an exemplary video encoder that may be used in a device of FIG. 1A or FIG. IB. Frames or part of frames from a video sequence may be placed in an input frame buffer 242 inside a video encoder 118 that may be part of CODEC 24.

WO 2007/038357 PCT/US2006/037130
An input frame from input frame buffer 242 may be parsed into blocks (the video blocks may be of any size, but standard square video block sizes are 4x4, 8x8, or 16x16) and sent to video block buffer 243. Video block buffer 243 typically sends a video block to subtractor 244. Subtractor 244 subtracts video block x from the output of switch 246. Switch 246 may switch between intra-coding and inter-coding prediction modes of encoding. If switch 246 is enabling an inter-coding prediction mode, then the resulting difference from x and a video block from a different (previous or subsequent) frame is compressed through texture encoder 247. If switch 246 enables an intra-coding prediction mode, then the resulting difference from x and predicted value from a previous video block in the same frame is compressed through texture encoder 247. [0038] Texture encoder 247 has a DCT block 248 which transforms the input x (the video block or difference block) from the pixel domain to a spatial frequency domain. In the spatial frequency domain, data is represented by DCT block coefficients. The .DCT block coefficients represent-the number and degree of the spatial frequencies detected in the video block. After a DCT is computed, the DCT block coefficients may be quantized by quantizer 250, in a process is known as "block quantization." Quantization of the DCT block coefficients (coming from either the video block or difference video block) removes part of the spatial redundancy from the block. During this "block quantization" process, further spatial redundancy may sometimes be removed by comparing the quantized DCT block coefficients to a threshold. This comparison may take place inside quantizer 250 or another comparator block (not shown). If the magnitude of a quantized DCT block coefficient is less than the threshold, the coefficient is discarded or set to a zero value.
[0039] After block quantization, the resulting output may be sent to two separate structures: (1) a texture decoder 265, and (2) an entropy encoder 255. Texture decoder 265 comprises a de-quantizer 266 which aids in the production of a reconstructed video block or frame; to be used with a coding prediction mode. The entropy encoder 255 produces a bitstream for transmission or storage. Entropy encoder 255 may contain a scanner 256 which receives the block quantized output and re-order it for more efficient encoding by variable length coder (VLC) 258. VLC 258 may employ the use of run-length and huffman coding techniques to produce an encoded bit-stream. The encoded bitstream is sent to output buffer 260. The bitstream may be sent to rate controller 262. While maintaining a base quality, rate controller 262 budgets the number of

WO 2007/038357

PCT/US2006/037130

quantization bits used by quantizer 250. Entropy encoding is considered a non-lossy form of compression. Non-lossy compression signifies that the data being encoded may be identically recovered if it is decoded by an entropy decoder without the encoded data having been corrupted. Entropy encoder 255 performs non-lossy compression.
[0040] Lossy compression means that as a result of the encoding, an input, x, will not produce an identical copy of x even though the encoded input has not been corrupted. The reconstructed input has "lost" part of its information. Texture encoder 247 performs lossy compression. A typical video encoder 118 usually has a local texture decoder 265 to aid in the compensation of both the inter-coding and intra-coding prediction modes. De-quantizer 266, inverse DCT 268, and the output of switch 246 that is sent to adder 269 work together to decode the output of texture encoder 247 and reconstruct the input x that went into texture encoder 247. The reconstructed input, y, looks similar to x but is not exactly x. A general video "decoder" typically comprises the functionality of the de-quantizer 266* inverse DGT 687 and the output of switch 246 that is sent to adder 269.
[0041] The reconstructed input may be sent to memory buffer 281. Inside memory buffer 281 there may be two memory buffers: (1) reconstructed new frame buffer 282; and (2) reconstructed old frame buffer 284. Reconstructed new frame buffer 282, stores the currently processed reconstructed frame (or partial frame). Reconstructed old frame buffer 284 stores a past processed reconstructed frame. The past processed reconstructed frame is used as a (reconstructed) reference frame. The reconstructed reference frame may be a frame that is before or after the current frame in input frame buffer 242. The current frame (or a video block from the current frame) or differences between the current frame and the reconstructed reference frame (or a video block from the difference block) is what is "currently" being encoded. After the current frame has finished encoding and before the next frame in input from input frame buffer 242 is fetched to be encoded, the reconstructed old frame buffer 284 is updated with a copy with the contents of the reconstructed new frame buffer 282.
[0042] Reconstructed new frame buffer 282 may send the reconstructed video block it received to be used in spatial predictor 286. Reconstructed old frame buffer 284 sends a past processed reconstructed video block to MEC (motion estimation and compensation block) 287. MEC block comprises motion estimator 88 and motion compensator 290. Motion estimator 288 generates motion vectors (MV) 92 and motion vector predictors

WO 2007/038357 PCT/US2006/037130
(MVP) 294 that may be used by motion compensator 290 to compensate for differences from other frames than the one being encoded. MVs 292 may also be used by entropy encoder 255. In some standards, such as ITU H.264, the output of spatial predictor 286 is used in intra-frame prediction mode and fed back both to subtracter 244 and adder 269. In some standards, such as MPEG-4 or JPEG, there is no spatial predictor 286. [0043] FIG. 3 illustrates two partial block rows of any image or frame. As an example, let block row N-l and N be block row 3 and block row 4. In block row 3 330 there are nine video blocks. For illustrative purposes 16x16 blocks may be used throughout the description of the disclosure. Hence, macroblocks (MB's), 331-339 are in row 3 330, and in row 4 340 there are nine MB's 341-749. The MB's are drawn showing both the block row number and the location relative to the Mth macroblock. M denotes a current macroblock. Typically block row 3 is processed before block row 4. Within this disclosure processing a block row means processing a row of macroblocks as outlined in row 3 330 and row 4340. In general, it can also mean the processing Of any size of video block row.
[0044] When calculating motion vectors relative to macroblock (MB) 345 in various standards, such as H.264, MPEG-4, and H.263, etc., it may be desired to have prior knowledge of MB modes and motion vectors of the neighboring MB 344, MB 335, and MB 336 (or MB 334 when MB 336 is not available). For example, in H.264, the INTER modes for a P-frame can be INTER 16x16, INTER 16x8, INTER 8x16, INTER 8x8. If it is an INTER 8x8 mode then a further partition can be made to select INTER 8x4, INTER 4x8 or INTER 4x4 modes. The mode is not only dependent on they type (INTER) but also the size. There can also be INTRA mode and SKIP mode. Although there are other conditions that invoke SKIP mode, one condition that may invoke SKIP mode is when the MV equals the MVP.
[0045] Motion estimation, for example, typically requires a larger amount of computational resources than any other process of video encoding. For this reason, it is highly desirable to perform motion estimation in a manner that can reduce computational complexity and also help in improving the compression ratio. The motion estimation techniques described herein may advance these goals by using a search scheme that performs the searching at multiple spatial resolutions, thereby reducing the computational complexity without any loss in accuracy. In addition, a cost function, also known as a distortion measure is proposed, that includes the cost of

WO 2007/038357 PCT/US2006/037130
encoding motion vectors. A motion estimator may also use multiple candidate locations of a search space to improve the accuracy of video encoding, and the search area around the multiple candidates may be programmable, thus making the process scalable with fame rate and picture sizes. Finally, a motion estimator may also combine cost functions for many small square video blocks, e.g., 4 by 4 video blocks, to obtain the cost for the various larger block shapes, e.g., 4 by 8 video blocks, 8 by 4 video blocks, 8 by 8 video blocks, 8 by 16 video blocks, 16 by 8 video blocks, 16 by 16 video blocks, and so forth..
For many operations and computations, a motion vector predictor (MVP) is used to add a cost-factor for motion vectors deviating from the motion vector predictor. The MVP may also provide an additional initial motion vector, which can be used to define searches, particularly at high resolution stages of a multi-stage search. The calculation of distortion measure values that depend at least in part the motion vector prediction values is part of the cost-factor. Distortion measure values may also help quantify the number of bits needed to encode different motion vectors.
[0046] FIG. 4 is an exemplary flow chart describing the process of how motion estimation can be done throughout a frame. First a setup procedure 400 is initiated. As part of the setup a reference (a past or future) frame is loaded 402 into a memory. For example, the memory may be a local memory or off-memory candidate RAM. Then a current frame is loaded 404 into a memory. For example, the memory may also be part of the local memory or off-memory encode RAM. This is followed by a selection of the current macroblock M in a block row. Then a search space is identified 406 around the current macroblock M. Once a search space has been identified from a reference frame 408, the act of processing a 2-D stage pipeline over two video block rows in a current frame 410 takes place. This is an improvement over the current art which only considers processing of one video block row at a time via a 1-D stage pipeline. After each 2-D stage pipeline act is performed, decision block 412 checks if the end of two block rows has been reached. If the end of the two block rows has not been reached (NO), then the search space in the reference frame is refreshed 414, and processing of the 2-D stage pipeline over two block rows in a current frame 410 continues. If the end of two block rows has been reached (YES) then another check is performed by decision block 416. Decision block 416 checks if it is the end of the last two block rows in the

WO 2007/038357 PCT/US2006/037130
current frame. If it is not the end of the last two block rows in a current frame (NO), then the act of incrementing to the next two block rows 418 is performed, and processing of the 2-D stage pipeline over two block rows in a current frame 410 continues. If is the end of the last two block rows in a current frame (YES) processing of the current frame ends.
[0047] Motion estimation (ME) involves processing a video block by at least two engines. The first engine is a fetch engine (FE), and may fetch video blocks from a memory to be encoded. The second engine uses a distortion metric to find a similar matching block which minimizes the distortion. The second engine is an integer search engine (ISE), and may employ a hierarchical search with search being performed in a finer resolution as it proceeds. There may also be a third engine, a fractional and spatial search engine (FSE), which may use a finer resolution search to find a similar matching block which minimizes the distortion. The FSE may or may not use the results of the second engine, as a starting point. For exemplary purposes,- description of the motion estimation technique is done with three engines, the fetch Engine (FE), the integer Search Engine (ISE), and the fractional and spatial search engine (FSE). These three engines process a video block in a sequential fashion. Typically, to minimize the delay of the processing, the three engines are pipelined over three stages. That is to say, an FE, ISE, and FSE all operate in parallel in a three stage pipeline. For example, during stage 1, an FE fetches from a memory the current macroblock M, while simultaneously, an ISE positioned with an integer pixel anchor point at the corner of a macroblock M-l tries to find the best match macroblock from a reference (a past or future) frame. Still during stage 1, simultaneously operating at the same time as the FE and ISE engines, an FSE positioned at a fractional pixel anchor point in macroblock M-2 tries to find the best match macroblock from a reference (a past or future) frame. It takes three stages to complete the complete processing of one video block to generate a motion vector. At the end of three stages there will have been three fetches, three integer searches performed and three fractional and spatial searches performed. Typically, the one dimensional (ID) three stage pipeline operates sequentially over one video block row. No processing takes place in the second block row until the whole video block row has been completely processed.
[0048] FIG 5 illustrates a conceptual illustration of the ID three stage pipeline. As can be seen most of row 3 330 is being processed. In the upper left comer of FIG 5, two of

WO 2007/038357 PCT/US2006/037130
the three engines are running during stage 1 501 of the 3, M-3 MB pipeline 520. FE operates on macroblock 3, M-3 332, and ISE operates on macroblock 3, M-4 331. Stage
2 502 of the 3, M-3 MB pipeline 520 and stage 3 503 of the 3, M-3 MB pipeline 520 complete the generation of motion vectors after the FSE is completed on the 3, M-3 MB (macroblock) 332. As mentioned previously, in general, at each stage a fetch, an integer search, and a fractional and spatial search is performed. With the exception of the initial fetch of the 3, M-4 macroblock 331, all macroblocks (MBs) in row 3 330 of FIG.
3 are processed in FIG. 5. For 1-D pipelines 521 - 527, stagel of the next pipeline is always stage 2 of the previous pipeline. Stage 2 of the next pipeline is stage 3 of the previous pipeline. Thus, stages 502-510 may be a stage 1, stage 2 or stage 3 stage in a pipeline, depending on which pipeline is currently operating. In general it takes N+2 stages to process N macroblocks (i.e. video blocks) using a ID pipeline.
[0049] For a given macroblock, the ID pipeline technique constrains the current ISE to not use the FSE result.of the previous motion block to its advantage; For example," when the FE is fetching the encode data and updating the memory for MB 335, ISE is working on integer search for MB 334 and FSE is performing fractional search and spatial estimation for MB 333. An inherent problem with the ID pipeline is that ISE does not know the modes and final motion vectors (MV's) of its left neighboring MB and thus is not able to obtain an accurate estimate of the motion vector predictors (MVP's). As a result the motion vector computation may be slightly off. Hence, by using a ID pipeline technique, there exists a mutual dependency between motion estimation (generating MV's) and mode decision (generating MVP's).
[0050] A motion estimation technique that resolves the mutual dependencies between motion estimation and mode decision is illustrated in FIG 6. FIG. 6 is an exemplary diagram illustrating a three stage two-dimensional (2D) pipeline, motion estimation technique that allows the ISE prior knowledge of all the relevant neighboring MB modes and MV's. In FIG. 6, the two video block rows are processed in a ping pong fashion. During the ping portion, two video blocks from a first video block row are processed, while simultaneous processing on another video block row is taking place. During the pong portion, two video blocks from a second video block row are processed, while on a first video block row another video block is being processed. [0051] In the upper left corner of FIG. 6, illustrates stage 1 601 of the 3, M-0 MB 2D pipeline 621. FE operates on macroblock 3, M-0 335, and ISE operates on macroblock

WO 2007/038357 PCT/US2006/037130

4, M-3 342, while FSE operates on macroblock 3, M-l 334. During stage 2 602 of the 3, M-0 MB 2D pipeline 621, FE operatres on macroblock 4,M-2 343, and ISE operates on macroblock 3, M-0 335, while FSE operates on macroblock 4, M-3 342. During stage 3 603 of the 3, M-0 MB 2D pipeline 621, FE operatres on macroblock 3.M+1 336, and ISE operates on macroblock 4, M-2 343, while FSE operates on macroblock 3, M-0 335. After completion of the 3, M-0 MB 2D pipeline 621 generation of motion vectors after the FSE is completed on the 3, M-0 MB 335. Stages 602-609 may be used as described above to complete 2D pipelines 622-627. Every "ping/pong", see FIG. 6, represents an FE/FSE operation on one row while there is an ISE on another row. The ping pong terminology is associated with the switching of roles between rows. For example, during the "ping" FE/FSE operate on row 3, and ISE operates on row 4. During the "pong" FE/FSE operate on row 4, and ISE operates on row 3.
[0052] An advantage of using the 2D pipeline is that two macroblocks from at least two different rows but close in proximity (in this example, horizontal position) may share a majority of the search area. Since the pong can re-use search area from the ping there will be less fetches required. Processing in this ping-pong fashion reduces the bandwidth on a communication bus. The number of refreshes to an external memory to update a search area may also significantly be reduced. The number of refreshes to an external memory to update a search area may also significantly be reduced.
[0053] In addition, macroblock 345 can use the mode decision from relevant neighboring macroblocks, 344, 335, and 334 (or 336 if 334 is not available). The use of mode decision means that the motion vector predictor from each of those relevant neighboring macroblocks may aid in producing an accurate estimate of the motion vector. Without the 2D ping pong pipeline, an accurate motion vector predictor from neighboring macroblock 344 may not be available in the computation of macroblock's 345 motion vector generation. Hence, the two-dimensional pipeline resolves a mutual dependency between a motion estimation video block and a mode decision.
[0054] A number of different embodiments have been described. The techniques may be capable of improving video encoding by improving motion estimation. The techniques may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the techniques may be directed to a computer readable medium comprising program code, that when executed in a device that encodes video sequences, performs one or more of the methods mentioned above. In that case,

WO 2007/038357 PCT/US2006/037130
the computer readable medium may comprise random access memory (RAM) such as synchronous dynamic random access memory (SDRAM), read-only memory (ROM), non-volatile random access memory (NVRAM), electrically erasable programmable read-only memory (EEPROM), FLASH memory, and the like. [0055] The program code may be stored on memory in the form of computer readable instructions. In that case, a processor such as a DSP may execute instructions stored in memory in order to carry out one or more of the techniques described herein. In some cases, the techniques may be executed by a DSP that invokes various hardware components such as a motion estimator to accelerate the encoding process. In other cases, the video encoder may be implemented as a microprocessor, one or more application specific integrated circuits (ASICs), one or more field programmable gate arrays (FPGAs), or some other hardware-software combination. These and other embodiments are within the scope of the following claims.

WO 2007/038357 PCT/US2006/037130
We claim:
1. A device having a motion estimator using a two dimensional pipeline to generate
accurate motion vectors, motion vector predictors, and mode decision for a current
video block, the two dimensional pipeline using completed motion vector, motion
vector predictors, and mode decision calculations of relevant neighboring video
blocks, including a preceding video block on a same video block row as the current
video block, prior to the generation of the accurate motion vector, motion vector
predictor, and mode decision of the current video block.
2. The device of claim 1, wherein the two dimensional pipeline comprises using a fetch engine and a fractional spatial search engine operating on a video block row, while an integer search engine operates on another video block row.
3. The device of claim 1, wherein the two dimensional pipeline involves processing at least two stages in the pipeline.
4. The device of claim 1, wherein the video block is a macroblock, an 8x8 block, an 8x4 block, a 4x8 block, a 4x4 block, or a 2x2 block.
5. In a device including a motion estimator, a method of performing two dimensional pipeline processing to generate accurate motion estimation parameters for a current video block, the two dimensional pipeline processing includes identifying previously calculated motion estimation parameters of neighboring video blocks, including a preceding video block on a same video block row as the current video block, prior to the generation of the accurate motion estimation parameters of the current video block.
6. The method of claim 5, wherein the motion estimation parameters are a motion vector from the current video block, motion vector predictors associated with the current video block, and mode decision..

WO 2007/038357 PCT/US2006/037130

7. The method of claim 5, wherein the two dimensional pipeline processing comprises using a fetch engine and a fractional spatial search engine operating on a video block row, while an integer search engine operates on another video block row.
8. The method of claim 5, wherein the two dimensional pipeline processing involves at least two stages in the pipeline.
9. The method of claim 5, wherein the video block is a macroblock, an 8x8 block, an 8x4 block, a 4x8 block, a 4x4 block, or a 2x2 block.

10. A device having a motion estimator including means for performing two dimensional pipeline processing to generate accurate motion estimation parameters for a current video block, the two dimensional pipeline processing having means for identifying previously, calculated motion estimation parameters of relevant neighboring video blocks, including a preceding video block on a same video block row as the current video block, prior to the generation of the accurate motion estimation parameters of the current video block.
11. The device of claim 10, wherein the motion estimation parameters are a motion vector from the current video block, motion vector predictors associated with the current video block, and mode decision.
12. The device of claim 10, wherein the two dimensional pipeline processing comprises using a fetch engine and a fractional spatial search engine operating on a video block row, while an integer search engine operates on another video block row.
13. The device of claim 10, wherein the two dimensional pipeline processing involves at least two stages in the pipeline.
14. The device of claim 10, wherein the video block is a macroblock, an 8x8 block, an 8x4 block, a 4x8 block, a 4x4 block, or a 2x2 block.

WO 2007/038357

PCT/US2006/037130

15. A computer program product comprising a computer readable medium having
instructions stored thereon when executed, generate a motion vector for a current video
block, the computer program product comprising sets of instructions for:
performing two dimensional pipeline processing to generate accurate motion estimation parameters for a current video block, the two dimensional pipeline processing further includes instructions for identifying previously calculated motion estimation parameters of neighboring video blocks, including a preceding video block on a same row as the current video block, prior to the generation of the accurate motion estimation parameters of the current video block.
16. The computer program product of claim 15, wherein the motion estimation parameters are a motion vector from the current video block, motion vector predictors associated with the current video block, and mode decision.
17. The computer program product of claim 15, wherein the two dimensional pipeline processing comprises using a fetch engine and a fractional spatial search engine operating on a video block row, while an integer search engine operates on another video block row.
18. The computer program product of claim 15, wherein the two dimensional pipeline processing involves at least two stages in the pipeline.
19. The computer program product of claim 15, wherein the video block is a macroblock, an 8x8 block, an 8x4 block, a 4x8 block, a 4x4 block, or a 2x2 block.
20. A method of video encoding comprising:
computing a motion vector predictor based on motion vectors previously calculated for video blocks in proximity to a current video block to be encoded; using the motion vector predictor in searching for a prediction block used to encode the current video block; and
calculation of at least one motion vector predictor and at least one motion vector by processing at least two video block rows using a two dimensional pipeline.

WO 2007/038357 PCT/US2006/037130
21. The method of claim 20, further comprising generating a difference block indicative of differences between the current video block to be encoded and the best prediction block.
22. The method of claim 21, further comprising identifying a motion vector to the prediction video block used to encode the current video block including calculating distortion measure values that depend at least in part the motion vector prediction value.
23. The method of claim 22, wherein the distortion measure values quantify number of bits needed to encode different motion vectors.
24. The method of claim 20, further comprising performing searches in stages at different spatial resolutions to identify the-motion vector to the best prediction block used to encode the current video block.

25. The device of claim 20, wherein the two dimensional pipeline comprises using a fetch engine and a fractional spatial search engine operating on a video block row, while an integer search engine operates on another vide block row.
26. A video encoding device comprising:
a motion estimator that identifies a motion vector to a best prediction block used to encode a current video block, including calculating distortion measure values that depend at least in part on an amount of data associated with different motion vectors;
a motion compensator that generates a difference block indicative of differences between the current video block to be encoded and the prediction video block; and
calculation of at least one motion vector predictor and at least one motion vector by using a two dimensional pipeline over at least two video block rows.

WO 2007/038357 PCT/US2006/037130

27. The video encoding device of claim 26, wherein the motion estimator performs searches in stages at different spatial resolutions to identify the motion vector to the best prediction block used to encode the current video block.
28. The video encoding device of claim 26, wherein the video encoding device computes a motion vector predictor based on motion vectors previously calculated for video blocks in proximity to the current video block to be encoded, wherein the motion vector predictor value defines a search in at least one of the stages and is also used to calculate the distortion measure values.

ABSTRACT
"MULTI-DIMENSIONAL NEIGHBORING BLOCK PREDICTION FOR VIDEO
ENCODING"
The motion estimation techniques and video encoding device(s) described use a two dimensional pipeline to generate accurate motion estimation parameters for a current video block. The two dimensional pipeline uses previously calculated motion estimation parameters of relevant neighboring video blocks, including a preceding video block on a same row as the current video block, prior to the generation of the accurate motion vectors, motion vector predictors, and mode decision of the current video block The use of the two dimensional pipeline allows accurate motion vector prediction from neighboring video blocks previously not available, in the computation of motion vectors. Three engines may be used in the two dimensional pipeline, a fetch engine, an integer search engine and a fractional and spatial search engine. While the fetch engine and fraction and spatial search engine operate on one row, the integer search engine operates on another row.

Documents:

755-MUMNP-2008-ABSTRACT(GRANTED)-(25-2-2013).pdf

755-mumnp-2008-abstract.doc

755-mumnp-2008-abstract.pdf

755-MUMNP-2008-CLAIMS(AMENDED)-(3-1-2013).pdf

755-MUMNP-2008-CLAIMS(GRANTED)-(25-2-2013).pdf

755-mumnp-2008-claims.doc

755-mumnp-2008-claims.pdf

755-MUMNP-2008-CORRESPONDENCE(10-1-2013).pdf

755-mumnp-2008-correspondence(12-5-2008).pdf

755-MUMNP-2008-CORRESPONDENCE(13-8-2012).pdf

755-MUMNP-2008-CORRESPONDENCE(22-10-2008).pdf

755-MUMNP-2008-CORRESPONDENCE(29-6-2012).pdf

755-MUMNP-2008-CORRESPONDENCE(IPO)-(25-2-2013).pdf

755-mumnp-2008-correspondence-others.pdf

755-mumnp-2008-correspondence-received.pdf

755-mumnp-2008-description (complete).pdf

755-MUMNP-2008-DESCRIPTION(GRANTED)-(25-2-2013).pdf

755-MUMNP-2008-DRAWING(3-1-2013).pdf

755-MUMNP-2008-DRAWING(GRANTED)-(25-2-2013).pdf

755-mumnp-2008-drawings.pdf

755-MUMNP-2008-EP DOCUMENT(3-1-2013).pdf

755-MUMNP-2008-FORM 1(13-8-2012).pdf

755-MUMNP-2008-FORM 1(3-1-2013).pdf

755-MUMNP-2008-FORM 13(13-8-2012).pdf

755-MUMNP-2008-FORM 2(GRANTED)-(25-2-2013).pdf

755-mumnp-2008-form 2(title page)-(17-4-2008).pdf

755-MUMNP-2008-FORM 2(TITLE PAGE)-(GRANTED)-(25-2-2013).pdf

755-MUMNP-2008-FORM 26(3-1-2013).pdf

755-MUMNP-2008-FORM 3(22-10-2008).pdf

755-MUMNP-2008-FORM 3(29-6-2012).pdf

755-MUMNP-2008-FORM PCT-IB-373(3-1-2013).pdf

755-MUMNP-2008-FORM PCT-ISA-237(3-1-2013).pdf

755-mumnp-2008-form-1.pdf

755-mumnp-2008-form-18.pdf

755-mumnp-2008-form-2.doc

755-mumnp-2008-form-2.pdf

755-mumnp-2008-form-26.pdf

755-mumnp-2008-form-3.pdf

755-mumnp-2008-form-5.pdf

755-mumnp-2008-form-pct-ib-304.pdf

755-mumnp-2008-pct-search report.pdf

755-MUMNP-2008-PETITION UNDER RULE-137(10-1-2013).pdf

755-MUMNP-2008-REPLY TO EXAMINATION REPORT(3-1-2013).pdf

755-MUMNP-2008-SPECIFICATION(AMENDED)-(3-1-2013).pdf

755-MUMNP-2008-SPECIFICATION(MARKED COPY)-(3-1-2013).pdf

755-MUMNP-2008-US DOCUMENT(3-1-2013).pdf

755-mumnp-2008-wo international publication report(17-4-2008).pdf

abstract1.jpg

« Previous Patent

Next Patent »

Patent Number

255476

Indian Patent Application Number

755/MUMNP/2008

PG Journal Number

09/2013

Publication Date

01-Mar-2013

Grant Date

25-Feb-2013

Date of Filing

17-Apr-2008

Name of Patentee

QUALCOMM INCORPORATED

Applicant Address

5775 MOREHOUSE DRIVE, SAN DIEGO, CALIFORNIA 92121-1714,

Inventors:

#	Inventor's Name	Inventor's Address
1	WANG, KAI	7565 CHARMANT DRIVE, #416, SAN DIEGO, CA 92122
2	MALAYATH, NARENDRANATH	10710 SABRE HILL DRIVE, #229, SAN DIEGO, CA 92128
3	NAGARAJ, RAGHAVENDRA C.	3415 LEBON DRIVE, #416, SAN DIEGO, CA 92122

PCT International Classification Number

H04N7/26

PCT International Application Number

PCT/US2006/037130

PCT International Filing date

2006-09-22

PCT Conventions:

#	PCT Application Number	Date of Convention	Priority Country
1	11/533,745	2006-09-20	U.S.A.
2	60/719,891	2005-09-22	U.S.A.