Title of Invention  METHOD AND DEVICE FOR POSTPROCESSING DIGITAL IMAGES 

Abstract  A method of postprocessing digital images comprising pixels, the method comprising the steps of: frequency transformation (21), suitable for delivering a set of transformed coefficients (Y) from a set of pixels (y), extraction (22) of original lowfrequency coefficients (Ybfo) and original highfrequency coefficients (Yhfo) comprised in the set of transformed coefficients, correction (23), suitable for delivering a set of corrected transformed coefficients (Yc) from the original lowfrequency coefficients, and combination (24), suitable for delivering a set of combined transformed coefficients (Yadd) from the original highfrequency coefficients and from the set of corrected transformed coefficients. 
Full Text  Method and device for postprocessing digital images FIELD OF THE INVENTION The invention relates to a method and a device for postprocessing digital images comprising pixels, The invention also relates to a computer program product for performing such a postprocessing method. The invention notably finds its application in the correction of digital images that have previously been encoded and subsequently decoded in accordance with a blockbased encoding technique, for example, the MPEG standard, so as to attenuate visual artifacts caused by the blockbased encoding technique, BACKGROUND OF THE INVENTION International patent application WO 97/29594 describes a method and a device for post"processing decoded video data so as to minimize the blocking artifacts in an image without affecting the contrast. To this end, the method of postprocessing data in accordance with the prior art, described in Fig. I, comprises the steps of: LPF filtering (I I) decoded video data (x), suitable for delivering filtered data (xf), DCT transform (12,13) of the filtered data and the decoded video data, suitable for delivering transformed filtered data (Xf) and transformed decoded data (X), adjusting ADJ (14) the lowfrequency coefficients comprised in the transformed filtered data, suitable for delivering adjusted lowfrequency coefficients (Xbf), combining (16) the highfiequency coefficients (Xhf) comprised in the Itansfonned decoded data and the adjusted lowfrequency coefficients, suitable for delivering combined transformed data (Xc), and IDCT transform (17) of the combined transformed data, suitable for delivering processed data (xc) that are ready for display on a screen. This method of postprocessing decoded video data necessarily comprises a lowpass filtering step for filtering the highfiequency components. This method allows extraction, in the spatial domain, of the lowfrequency data on which a first DCT transform and an adjustment are applied in order to deliver adjusted lowfrequency coefficients. The highfrequency coefficients are extracted (15) by way of a second DCT transform of the decoded video data. Thus, one part of the dataprocessing method is realized in the spatial domain, which renders the method both complex to be carried out and costly in terms of computing. Moreover, a lowfrequency coefficient comprised in the transfomied filtered data is adjusted vnthin the interval [Xqq/2,Xq+q/2], where Xq is the value of said quantized lowfrequency coefficient and q is the value of the quantization step, so that such a method necessitates access to the encoding parameters, which is not always possible. OBJECT AND SUMMARY OF THE INVENTION It is an object of the present invention to provide a method of postprocessing decoded video data, which can be performed in a simple and economical manner. To this end, the method of postprocessing digital images according to the invention is characterized in that it comprises the steps of: frequency transformation, suitable for delivering a set of transformed coefficients from a set of pixels, extraction of original lowfrequency coefficients and original highfrequency coefficients comprised in the set of transformed coefficients, correction, suitable for delivering a set of corrected transformed coefficients from the original lowfrequency coefficients, and combination, suitable for delivering a set of combined transformed coefficients from the original highfrequency coefficients and from the set of corrected transformed coefficients. In such a postprocessing method, data are processed, and particularly the original high and lowfrequency coefficients are separated, in the frequency domain. The postprocessing method can therefore be carried out in a simpler manner because the original lowfrequency coefficients are solely extracted and collected via the transfonned coefficients. Consequently, neither the step of lowpass filtering in the spatial domain, nor a double frequency transformation of the filtered data and the decoded video data is necessary. Said method is particularly adapted if the frequency transformation step uses the same type of frequency transformation as that used in the technique of blockencoding during a previous encoding of video data, for example, a DCT transform in the case of data that have previously been encoded and subsequently decoded in accordance with the MPEG standard. It also allows better control of the correction of blocking artifacts. In addition, ihe postprocessing method is more effective, particularly when a first half of the set of transformed coefficients constitutes the original lowfrequency coefficients and a second half of said set constitutes the original highfrequency coefficients. It is also a very fle?dble method as regards the restitution of the high frequencies in the image, wherein the combination step can linearly combine corrected transformed highfrequency coefficients and original highfrequency coefficients so as to deliver the set of combined transformed coefficients. BRIEF DESCRIPTION OF THE DRAWINGS These and other aspects of the invention are apparent from and will be elucidated, by way of nonlimitative example, with reference to the embodiments described hereinafter. In the drawings: Fig. 1 illustrates the method of postprocessing data in accordance with the prior art. Fig. 2 is a diagram showing the principal steps of the method of post¬processing data according to the invention. Fig. 3 illustrates a method of correcting blocking artifacts, and Fig. 4 is a diagram showing an embodiment of the method of postprocessing data according to the invention, comprising such a step of correcting blocking artifacts. DESCRIPTION OF PREFERRED EMBODIMENTS The present invention relates to a method of postprocessing digital images that have previously been encoded and subsequently decoded in accordance with a blockbased encoding technique and therefore comprise blocking artifacts. As will be described hereinafter, the method of postprocessing video data according to the invention is intended to determine: the original highfrequency coefficients that must be preserved within a block of transformed coefficients so as to maintain the det^ls of the image, such as contours, as well as the original lowfrequency coefficients that must be corrected so as to effectively suppress the blocking artifacts. iig. 2. snows diagraramatically the principal steps of the method of post¬processing decoded video data according to the invention. This method comprises the steps of: frequency transformation TF (21), suitable for delivering a set of transformed coefficients (Y) from a set of pixels (y), extraction SEP (22) of original lowfrequency coefficients (Ybfo) and original highfrequency coefficients (Yhfo) comprised in the set of transformed coefficients, correction COR (23), suitable for delivering a set of corrected transformed coefficients (Ye) fivsm the original lowfrequency coefficients, combination (24), suitable for delivering a set of combined transformed coefficients (Yadd) from the original highfrequency coefficients and from the set of corrected transformed coefficients, and inverse frequency transformation ITF (25) of the set of combined transformed coefficients, suitable for delivering a processed set of pixels (yout) that are ready for display on a screen. The set ofpixels is preferably a segment ofN pixels, with N = 8 in the case of the MPEG standard for which the encoding blocks generally comprise 8 rows of 8 pixels. The set of pixels may be alternatively constituted by either an entire or a partial encoding block. The frequency transformation step preferably uses a transformation of the DCT type which is particularly adapted to the MPEG standard. In the preferred embodiment, a segment of pixels is transformed in the transformation step into a segment of DCT coefficients. Subsequently, a first half of the segment of DCT coefficients, i.e. the first 4 DCT coefficients constituting the original lowfrequency coefficients (Ybfo), and a second half of the segment of DCT coefficients, i.e. the last 4 DCT coefficients constituting the original highfrequency coefficients (Yhfo), are extracted in the extraction step. Such a separation into two parts of the segment of DCT coefficients allows a better correction of blocking artifacts. It may also be easily adjusted optimally as compared with the priorart technique, which necessitates the optimal adjustment of a lowpass filter from an infinite number of available filters. The extraction step (22) also comprises a truncation substep during which the first 4 DCT coefficients are completed by 4 zero coefficients so as to deliver a segment of truncated DCT coefficients. This truncation substep is in a way similar to a lowpass filtering operation. In the correction step (23), the segments of truncated coefficients are then corrected so as to deliver a segment of corrected DCT coefficients comprising 4 corrected lowfrequency DCT coefficients (Ybfc) and 4 corrected highfrequency DCT coefficients (Yhfc). In the combination step (24), the 4 original highfrequency coefficients and the segment of corrected DCT coefficients are combined so as to deliver a segment of combined DCT coefficients (Yadd). In a particularly advantageous embodiment, the segment of combined DCT coefficients corresponds to the concatenation of the 4 corrected lowfrequency DCT coefficients (Ybfc) and the 4 original highfrequency coefficients (Yhfo). In the preferred embodiment, the segment of combined DCT coefficients corresponds to the concatenation of the 4 corrected lowfi^quency DCT coefficients (Ybfc) and the 4 combined highfrequency DCT coefficients, resulting in a linear combination of the 4 corrected highfrequency DCT coefficients (Yhfc) and the 4 original highfrequency coefficients (Yhfo), namely: Yhfadd = a. Yhfc + (lb).Yhfo where a and b are real values between 0 and I, with, for example, a=l/2 and b=a/4if Yhfc differs from 0 and b=l/2 if Yhfc is equal to 0. Such a combination gives more weight to the original highfrequency coefficients, if high frequencies appear in the corrected DCT coefficients, and infroduces an attenuation in the opposite case. The segment of combined DCT coefficients is finally transformed in the spatial domain via an IDCT transform, which delivers a segment of processed pixels (yout) for display on a screen. The method of postprocessing data also comprises at least a horizontal processing of an image, associated with at least a vertical processing of said image. Indeed, the blocking artifacts may be present at the borders of an encoding block, that is, in the four segments delimiting the block vertically or horizontally. If the image is processed horizontally, vertical blocking artifacts will be corrected; conversely, horizontal blocking artifacts vn\l be corrected if the image is processed vertically. The method of postprocessing data is successively applied to each of the two fields constituting a frame if the image is composed of two fields. It is preferably applied to luminance data in the digital image. The correction step which is carried out is based on the numerous blocking artifact correction methods that are known to those skilled in the art, and preferably those methods that do not use decoding parameters, which parameters are not always accessible. In the preferred embodiment, the data correction method is referred to as DFD method (DCT Frequency Deblocking). Such a data correction method comprises the following steps, shown in Fig. 3, namely: a step ofcomputing a first discrete cosine transform DCTl (31) of a first segment (u) of N pixels, with N = 8 in the example used, resulting in a first transformed segment U, a step ofcomputing a first discrete cosine transform DCTl (32) of a second segment (v) of N pixels, which second segment is adjacent to the fu"St, resulting in a second transformed segment V, a step of determming (33) a predicted maximum frequency (kpred) as a function of the maximum frequencies ku and kv of U and V, as follows: kpred = 2.max(ku,kv) + 2 with ku = max(ke {0,.. .,N1} / U(k) * 0), kv  max(ke {0... .,N1) / V(k) * 0), and max is the function giving the maximum of k among a set of determined values, a step of processing (35) a concatenated segment (w) comprising 2N pixels, i.e. 16 pixels in our case, and corresponding to the concatenation (34) of the first (u) and the second (v) segment, this processing step comprising the following substeps of: computing a second discrete cosine transform DCT2 (36) of the concatenated segment (w), resulting in a transformed concatenated segment W, correction (37) by setting those transformed data W(() to zero that have an odd frequency k which is higher than the predicted maximum frequency (kpred), delivering a corrected ttaiaformed concatenated segment Wc, computing an inverse discrete cosine transform IDCT2 (38) of the corrected transformed segment Wc, delivering a corrected concatenated segment (cw). In a preferred embodiment of the invention, filtering thresholds are introduced in accordance with the following principle: kumax  max( ke {0,.. .,N1} / abs(U(k)) > T ) kvmax = max( ke {0,...,N1} / abs(V(k)) > T ) where T is a threshold different from zero. In the determination step (33), a more precise predicted maximum frequency (kpred) is computed from the introduction of the threshold T, which allows a more effective correction of the blocking artifacts. The value of the threshold T is a function of the size of the segments u and v. Indeed, a part solely consisting of pixels of the segments u and v may be processed, for example, the pixels of either the even or the odd rows. The correction step COR (37) preferably comprises a substep of detecting natural contours from the values of the pixels of the initial segments u and v and the transformed segments U and V. This substep allows distinction of the natural contours of the blocking artifacts. A natural contour is detected if two conditions are met: the average values of the pixels of the segments u and v on both sides of a border between blocks differ by a considerable value which is higher than a predetermined threshold M, the segments u and v have a low spatial activity, which becomes manifest by the fact that the values ku and kv are small and less than a predetermined value kO. Fig. 4 is a diagram showing an embodiment of the method of postprocessing data according to the invention, including the DFD method of correcting blocking artifacts as described with reference to Fig. 3. Such a method of postprocessing video data comprises the steps of: DCT transform (41), suitable for delivering a first segment of DCT coefficients (U) from a first set of pixels (u), DCT transform (42), suitable for delivering a second segment of DCT coefficients (V) from a second set of pixels (v) adjacent to the first segment of pixels, extracting (43) first original lowftequency coefficients (Ubfo) and fust original highfi^quency coefficients (Uhfo) comprised in the first segment of DCT coefficients (U), suitable for delivering a first segment of truncated DCT coefficients (Ut) comprising the first original lowfi«quency coefficients, extracting (44) second original lowfrequency coefficients (Vbfo) and second Driginal highfrequency coefficients (Vhfo) comprised in the second segment of DCT loefficients (V), suitable for delivering a second segment of truncated DCT coefficients (Vt) :omprising the second original lowfrequency coefficients, correction (23), suitable for delivering a first and a second segment of ;orrected DCT coefficients (Uc,Vc, respectively) from the first and second original lowTcquency coefficients (Ubfo,Vbfo, respectively), comprising the substeps of: IDCT transform (231) of the first segment of truncated DCT loefficients (Ut), intended to deliver a first segment of a preprocessed pixel (ut), IDCT transform (232) of the second segment of truncated DCT ;oefficients (Vt), intended to deliver a second segment of a preprocessed pixel (vt). DFD correction (230) of the segments of preprocessed pixels in accordance with the principle described with reference to Fig, 3, suitable for delivering segments of corrected pixels (uc,vc), DCT transform (233) of the first segment of corrected pixels (uc), suitable for delivering a first segment of corrected DCT coefficients (Uc), DCT transform (234) of the second segment of corrected pixels (vc), suitable for delivering a second segment of corrected DCT coefficients (Vc), combination (45), suitable for delivering a first segment of combined DCT coefficients (Uadd) from the first original highfiequency coefficients (Uhfo) and the first segment of corrected DCT coefficients (Uc), combination (46), suitable for delivering a second segment of combined DCT coefficients (Vadd) from the second original highfrequency coefficients (Vhfo) and the second segment of corrected DCT coefficients (Vc), IDCT transform (47) of the first segment of combined DCT coefficients (Uadd), suitable for delivering a first segment of processed pixels (uout) for display on a screen, and IDCT transform (48) of the second segment of combined DCT coefficients (Vadd), suitable for delivering a second segment of processed pixels (vout), adjacent to the first segment of processed pixels (uout), for display on a screen. The postprocessing method described with reference to Fig. 4 has the advantage that it does not degrade the quality of the images that do not originally comprise any blocking artifacts. Indeed, in the presence of such images, the original lowfrequency coefficients are not subjected to any modification because the DFD collection method used and the combined highfrequency DCT coefficients remain identical to the original highfrequency coefficients. In the priorart technique, the use of a lowpass filter in the spatial domain may, on the other hand, lead to a degradation of the quality of an image that does not originally comprise any blocking artifact. The invention may be realized in the form of software embedded on one or several circuits performing the method of postprocessing data described hereinbefore, or in the form of items of hard ware. A device for postprocessing digital images, corresponding to saidmethod, is represented again by the functional blocks of Fig. 2. It comprises: frequency transformation means TF (21), suitable for delivering a set of decoded transformed coefficients (Y) from a set of pixels (y). means SEP (22) for extracting original lowfrequency coefficients (Ybfo) and original highfrequency coefficients (Yhfo) comprised in the set of transformed coefficients, correction means COR (23), suitable for delivering a set of conected transformed coefficients (Yc) from the original lowfrequency coefficients, and combination means (24), suitable for delivering a set of combined transformed coefficients (Yadd) from the original highfrequency coefficients and from the set of conected transformed coefficients, and means ITF (25) for inverse frequency fransformation of the set of combined transformed coefficients, suitable for delivering a set of processed pixels (yout) that are ready for display on a screen. Such a postprocessing device can be inserted between a video decoder and a television receiver in order to postprocess decoded digital images and to display the postprocessed digital images on the television receiver. Such a device can be built independently. It can also be part of the video decoder or of the television receivtt. There are numerous ways of implementing the previously described functions by means of software. In this respect. Figs. 2 to 4 are very diagrammatic. Therefore, although a Figure shows different functions in the form of separate blocks, it does not exclude the fact that a single piece of software performs several functions. This neither excludes the fact that a function can be performed by a set of software. It is possible to implement these functions by means of a suitably programmed video decoder circuit, a set top box, or a television receiver. A set of instructions comprised in a programming memory may cause the circuit to perform the different operations described hereinbefore with reference to Figs. 2 and 4. The set of instructions may also be loaded into the programming memory by reading a record carrier such as, for example, a disc comprising the set of instructions. Reading may alternatively be effected through a communication network such as, for example, the Internet. In this case, a service provider will put the set of instructions at the disposal of those interested. Any reference sign placed between parentheses in this text shall not be construed as limiting the claim. Use of the verb "comprise" and its conjugations does not exclude the presence of elements or steps other than those stated in the claims. Use of the article "a" or "an" preceding an element or a step does not exclude the presence of a plurality of such elements or steps. WE CLAIM: 1 A method of postprocessing digital images comprising pixels, the method comprising the steps of: frequency transformation (21), suitable for delivering a set of transformed coefficients (Y) from a set of pixels (y), extraction (22) of original lowfrequency coefficients (Ybfo) and original highfrequency coefficients (Yhfo) comprised in the set of transformed coefficients, correction (23), suitable for delivering a set of corrected transformed coefficients (Yc) from the original lowfrequency coefficients, and combination (24), suitable for delivering a set of combined transformed coefficients (Yadd) from the original highfrequency coefficients and from the set of corrected transformed coefficients. 2. The method of postprocessing digital images as claimed in claim 1, wherein the extraction step (22) is suitable for extracting a first half of the set of transformed coefficients (Y) as being the original lowfrequency coefficients (Ybfo) and a second half of the set of transformed coefficients as being the original highfrequency coefficients (Yhfo). 3. The method of postprocessing digital images as claimed in claim 1, wherein the combination step (24) is suitable for linearly combining corrected transfonned highfrequency coefficients (Yhfc) and original highfrequency coefficients (Yhfo) so as to deliver the set of combined transformed coefficients (Yadd). 4. A device for postprocessing digital images comprising pixels, the device comprising: frequency transformation means (21), suitable for delivering a set of decoded transformed coefficients (Y) from a set of pixels (y). means (22) for extracting original lowfrequency coefficients (Ybfo) and original highfrequency coefficients (Yhfo) comprised in the set of transformed coefficients, correction means (23), suitable for delivering a set of corrected fransformed coefficients (Yc) from the original lowfrequency coefficients, and combination means (24), suitable for delivering a set of combined transformed coefficients (Yadd) from the original highfrequency coefficients and from the set of corrected transformed coefficients. 5. A television receiver adapted to receive digital images and comprising a postprocessing device as claimed as in claim 4, suitable for postprocessing the digital picture in order to display postprocessed digital pictures on a screen of the television receiver. 

inpct20022028che abstract.pdf
inpct20022028che claims duplicate.pdf
inpct20022028che claims.pdf
inpct20022028che correspondence others.pdf
inpct20022028che correspondence po.pdf
inpct20022028che description (complete) duplicate.pdf
inpct20022028che description (complete).pdf
inpct20022028che drawings duplicate.pdf
inpct20022028che drawings.pdf
inpct20022028che form1.pdf
inpct20022028che form18.pdf
inpct20022028che form26.pdf
inpct20022028che form3.pdf
inpct20022028che form5.pdf
inpct20022028che others.pdf
inpct20022028che pct search report.pdf
inpct20022028che petition.pdf
Patent Number  228217  

Indian Patent Application Number  IN/PCT/2002/2028/CHE  
PG Journal Number  10/2009  
Publication Date  06Mar2009  
Grant Date  28Jan2009  
Date of Filing  09Dec2002  
Name of Patentee  KONINKLIJKE PHILIPS ELECTRONICS N.V  
Applicant Address  GROENEWOUDSEWEG 1, NL5621 BA EINDHOVEN,  
Inventors:


PCT International Classification Number  G06T 5/10  
PCT International Application Number  PCT/IB02/01250  
PCT International Filing date  20020408  
PCT Conventions:
