Title of Invention

"AN INTERPOLATION METHOD FOR A BINARY IMAGE"

Abstract An improved interpolation method in which a threshold value used for determining a pixel value of a pixel generated by interpolation according to a context (state value of adjacent pixels). In the interpolation method, the ambiguity between the interpolation value and the threshold value is removed by using the context (state value of the reference pixels around the interpolated pixel), thereby reducing the blocking and smoothing phenomena in the restored binary image.
Full Text BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to an interpolation method for a binary image, and more particularly, to an improved interpolation method in which a variable threshold value used for determining a pixel value to be generated by interpolation is determined according to a context (the state value of adjacent pixels) .
2. Description of the Related Art
Recently, a function for processing shape information has been added to MPEG-4. The shape information, which refers to object information of the image, is expressed as a binary image. To code such a binary image, MPEG-4 adopts a content-based arithmetic encoder (CAE). For lossy shape coding a down sampling method and an upsampling method are performed in each shape macro block. A shape image is divided into shape blocks which have MxM block size.
The down sampling refers to a method for reducing the binary image block according to a given conversion ratio. The reduced image block is transmitted together with the conversion ratio. Here, the conversion ratio is determined such that an error between the original binary image block and an binary image block restored later is within a predetermined range.
The reduced image block obtained by the down sampling is
1A

coded by a context based arithmetic encoder (CAE) and then transmitted.
The up sampling method is used to restore such a reduced image block. Up sampling is for restoring the reduced image block into the original binary image, by interpolation.
In such an up sampling process, an effective interpolation method must not cause excessive blocking and smoothing effects in the restored binary image.
SUMMARY OF THE INVENTION
To satisfy the above requirement, it. is an object of the present invention to provide an improved interpolation method in which a context (state value of pixels adjacent to a pixel (interpolated pixel) generated by interpolation) is used for interpolation, thereby reducing blocking and smoothing effects.
To achieve the above object, there is provided an interpolation method for a binary image, for restoring a reduced binary image, reduced from an original binary image by down sampling, into the original binary image, the method comprising the steps of: (a) calculating an interpolation value based on the pixel values (object pixel values) of the reduced image around an interpolated pixel; (b) calculating a context Cp(a state value of pixels (reference pixels) around the object pixels); (c) obtaining a threshold value corresponding to the calculated context; and (d) comparing the interpolation value with the threshold value of the step (c),
2

and setting the pixel value of the interpolated pixel as "1" if the interpolation value is greater than the threshold value, and setting the pixel value of the interpolated pixel as "0" if the interpolation value is equal to or less than the threshold value.
BRIEF DESCRIPTION OF THE ACCOMPANYING DRAWINGS
The above object and advantages of the present invention will become more apparent by describing in detail a preferred embodiment thereof with reference to the attached drawings in which:
FIG. 1A is a diagram illustrating coding and decoding methods for a binary image according to MPEG-4;
FIG. IB is a diagram illustrating a down sampling step shown in FIG. 1A;
FIG. 1C is a diagram illustrating an. up sampling step shown in FIG. 1A;
FIGs. 2A through 2D are diagrams illustrating an interpolation method according to the present invention;
FIG. 3 is a flowchart illustrating a threshold value learning method according to the present invention; and
FIGs. 4A through 4C show the effect of the interpolation method according to the present invention, in contrast to the conventional method.
DESCRIPTION OF THE PREFERRED EMBODIMENT In FIG. 1A, coding and decoding methods for a binary
3

image having object information according to the MPEG-4 includes a down sampling step 100, a coding step 102, a reverse coding step 104 and an up sampling step 106.
In the down sampling step 100, an Mx N binary image block is converted into an (MxCR)x(NxCR) binary image block. Here, CR is a convention ratio indicating the ratio in size of the reduced image block obtained by the down sampling with respect to the original binary image block.
In the coding step 102, the reduced image is encoded. In order to code the binary image including objects, MPEG-4 adopts a context based arithmetic encoder The reduced image coded through the coding step 102 is transmitted via a transmission route. In the reverse coding step 104, the coded reduced image is restored into the reduced image. In the up sampling step 106, the interpolation is performed on the reduced image to obtain the original binary image.
FIG. IB illustrates in detail the down sampling step 100 shown in FIG. 1A. The binary image 200 is made into a plurality of macro blocks, as shown in FIG. IB. Pixels indicated by a character "0" in a macro block 201 are converted into one pixel indicated by a character "X". Here, the size of the macro block is determined depending on the
4

conversion ratio. The conversion ratio of MPEG-4 may be 1, ½ or 1/4. Such conversion is performed on all macro blocks, resulting in an image reduced by the conversion ratio. In FIG. 1B, four pixels (indicated by "0") of a macro block are reduced to a pixel, that is, the convention ratio is equal to 1/2.
In the down sampling step, if half or more pixels of each group of four pixels in a macro block are equal to "1", the pixel value of the converted pixel becomes "1". Here, pixels having value "1" represent a part of the picture containing an object, and pixels having value "0" represent a part of the picture without an object. The boundary between pixels having value "1" and pixels having value "0" represents the boundary of the object information.
FIG. 10 illustrates in detail the up sampling step 106 shown in FIG. 1A. During the up sampling step, each pixel of the original binary image is restored by the interpolation using four pixels of the reduced image.
Referring to FIG. 1C, a conventional interpolation method will be described in detail.
In FIG. 1C, pixels A, B, C and D represent object pixels, which are pixels concerned in the interpolation, and pixels P1, P2, P3 and P4 represent interpolated pixels, which are pixels to be obtained by the interpolation. Here, the object pixels A, B, C and D surround the interpolated pixels P1, P2, P3 and P4. Also, the object pixels belong to the reduced image while the interpolated pixels belong to the restored
5

binary image.
First, an interpolation value INP[p] is obtained by a linear interpolation method as follows, using the object pixels A, B, C and D:

where, r and s are weights,"and r is greater than s. That is, for each interpolated pixel, the larger weight is applied to the pixel which is the closest to each interpolated pixel.
Next, the obtained interpolation value INP [P] and a threshold value THR are compared. Here, the threshold value THR is set to a value which is half of the largest possible interpolation value. If the interpolation value INP[P] is greater than the threshold value THR, the pixel value of that interpolated pixel becomes "1". Otherwise, the pixel value of the interpolated pixel becomes "0".
Here, only one threshold THR is applied while a plurality of interpolation values INP may be applied. Thus, if the interpolation value INP[P] and the threshold value THR are close to each other, it is unclear whether the pixel value of the interpolated pixel is correct or not. Accordingly, serious blocking or smoothing phenomenon is present in the restored binary image.
6

To reduce the blocking or smoothing, a threshold value to be compared with the interpolation value INP[P] is adaptively determined according to the context (state values of the pixels (reference pixels) around the interpolated pixel), thereby reducing ambiguity in determination of the pixel value of the interpolated pixel. Also, proposed threshold values are determined using a learning method.
An improved interpolation method according to the present invention will be described with reference to FIGs. 2A through 2D. In FIGs. 2A through 2D, pixels A, B, C, D, CO, C1, C2, C3 and C4 indicated by "0" are pixels of the reduced image, and pixels P1, P2, P3 and P4 indicated by "X" represent the interpolated pixels of the restored binary image.
First, the interpolation value of a pixel to be generated by the interpolation is calculated. Here, the interpolation value is calculated by a bilinear interpolation method. However, the interpolation value may alternatively be calculated by using the average of the object pixels or other methods.
The interpolation value is calculated using the bilinear interpolation method, by the following equation (1):
INP[P] = A- (1-h) (1-v) + B- (l-h)v + C.h(l-v) + D.hv ... (1)
where h and v of each term represent horizontal and vertical distances between the object pixel and the interpolated pixel included in the corresponding term. The interpolation value
7

INP[P] is determined as a value between 0 and 1.
The number of combinations of the object pixels is equal to 16, and the number of interpolation values INP[P] is 16. However, if duplicate values are not counted, the actual number of interpolation values is 12, which are all between 0 and 1. Accordingly, an integer t has the range of 0-11, and the number of proposed threshold values THRc[t] is equal to 12, including 0, 0.0625, 0.1875, 0.25, 0.375, 0.4375, 0.5625, 0.625, 0.75, 0.8125, 0.9375 and 1.
Then, the context (state value of the reference pixels) is calculated by the follow'ing equation (2) :

where P represents the position of the interpolated pixel, R represents the reference pixel, and k is an index of reference pixels and also a weight. Here, the index k is varied depending on the position of the interpolated pixel. In the present invention, five pixels around the object pixels are used as the reference pixels.
FIGs. 2A through 2D show positions and indices of the reference pixels used for interpolating the interpolated pixels P1 through P4, respectively.
The proposed threshold values THRc[t] are determined by a learning method, described below.
8

Then, the interpolation value INP [P] and the threshold value THRc based on the context are compared to determine the pixel value of the interpolated pixel.
If the interpolation value INP[P] is greater than the threshold value THRc, the pixel value of the interpolated pixel becomes "1". Otherwise, the pixel value of the interpolated pixel becomes "O".
Next, a method for determining the threshold value THRc based on the context of the reference pixel will be described. The threshold value THRc is obtained by the learning method. The learning is performed by the step of comparing the original binary image with the restored binary image.
In detail, referring to FIG. 3, first, arrays h[c][t] and THRc[t] are defined (step 310).
Here, the array h[c][t] holds numbers of hits, where a hit is the name given to when the restored pixel value is equal to the original pixel value. The hits are. determined by applying all of the possible threshold values to each context. Here, c is an index of the interpolation values, and t is an index of the proposed threshold values. In the case of FIG. 2, since the number of interpolation values and the number of proposed threshold values are both equal to 12, the size of the array becomes h[ll] [11] .
Also, the array THRc[t] stores the proposed threshold values. Here, t is an index of the proposed threshold values.
Then, the array h[c][t] is initialized to zero {step 320) .
9

After initializing the array hlc](t] to zero, the original binary image and the reduced image are input (step 330) .
Then, the location of an initial interpolated pixel is determined. The interpolation is performed from the upper-left to the lower-right by a raster scanning (step 340).
Once the location of the interpolated pixel, is determined, context is calculated by the above equation (2) (step 350).
The threshold index t is initialized (step 355).
Then, the interpolation value INP[P] of the interpolated pixel is calculated (step 360).
The current proposed threshold value from THRc[t] is compared with the calculated interpolation value INP[P] (step 370). If the interpolation value INP[P] is greater than the proposed threshold value THRc[t], the pixel value P of the interpolated pixel is set to "l" (step 372). Otherwise, the pixel value P of the interpolated pixel is set to "0" (step 374) .
The pixel value P of the interpolated pixel is compared with the original pixel value (step 380}. If the pixel value P is equal to the original pixel value, the corresponding value in the array h[c][t] is increased by "1". This value represents the number of hits (step 385). Otherwise, the step 385 is not performed.
It is checked whether the comparison of step 380 has been performed on all proposed threshold values of the current
10

interpolation value (step 390).
If the condition of step 390 is satisfied, it is determined whether the interpolation has been performed on all pixels of the binary image (step 400)
If the condition of step 390 is not satisfied, then the process returns to step 370.
If the interpolation has been performed on all pixels of the binary image, the proposed threshold value providing the highest frequency of hits is set as the threshold value of the corresponding interpolation value (step 400).
If the condition of step 390 is not satisfied, the process returns to step 350.
According to the interpolation method of the present invention, ambiguity of the restoration performed by using only the interpolation value obtained by the bilinear interpolation can be removed by using the context, thereby reducing restoration error in the restored binary image.
FIGs. 4A through 4C show the effect of the . interpolation method according to the present invention, in contrast to the conventional method. In detail, FIG. 4A shows the original binary image containing the MPEG-4 logo, FIG. 4B shows the result of the conventional interpolation, described with reference to FIG. 1C, and FIG. 4C shows the result of the interpolation according to the present invention. As can be seen from FIGs. 4B and 4C, the blocking and smoothing effects are sharply reduced compared to the conventional method.
As described above, in the interpolation method of the
11

present invention, the ambiguity in the comparison between the interpolation value and the threshold value is removed by using the context (state value of the reference pixels around the interpolated pixel), thereby reducing the blocking and smoothing phenomena in the restored binary image.
12

WE CLAIM:
1. An interpolation method for a binary image, for restoring a reduced binary image, reduced, from an original binary image by down sampling, into the original binary image, the method comprising the steps of:
(a) calculating an interpolation value based on the pixel
values of pixels (object pixels) around an interpolated pixel;
(b) calculating a context C,, (a state value of pixels
(reference pixels) around the object pixels);
(c) obtaining a threshold value corresponding to the calculated context; and
(d) comparing the interpolation value with the threshold value of the step (c), and setting the pixel value of the interpolated pixel as "1" if the interpolation value is greater than the threshold value, and setting the pixel value of the interpolated pixel as "0" if the interpolation value is equal to or less than the threshold value.

2. The interpolation method of claim 1, wherein the step (a) is calculated by a bilinear interpolation method.
3. The interpolation method of claim 1, wherein the step (b) is performed using the following equation:

13

where P repersents the position of the interpolated pixel, R represents the reference pixel, and k is an index of the reference pixel.
4. The interpolation method of claim 1, wherein the threshold value obtained in the step 5. The interpolation method of claim 4, wherein the proposed threshold values are obtained by the steps of:
(al) defining arrays h[cj [t] and THRc[t], where c is an index of the interpolation values, and t is an index of the proposed threshold values;
(a2) initializing the array h[cj [t] to zero, and receiving the original binary image and the reduced image;
(a3) determining the location of the interpolated pixel to be interpolated, according to a raster scanning, and calculating a context of the reference pixels;
(a4) initializing the threshold index t;
(a6) comparing the current proposed threshold value THRc[t] with the calculated interpolation value INP[P], and setting the pixel value of the interpolated pixel to "1" if the interpolation value INP[P] is greater than the proposed
14

threshold value THRc[t], and setting the pixel value of the interpolated pixel to "0" if the interpolation value INP[P] is less than or equal to the proposed threshold value THRc[t];
(a7) comparing the pixel value of the interpolated pixel with the original pixel value, and increasing the value of the array h[c][t] and the threshold index t by "1" if there is a hit, where a hit is when the pixel value of the interpolated pixel is equal to the original pixel value;
(a8) checking whether the comparison of the step a6 has been performed on all the proposed threshold values, and returning to the step a6 if the comparison has not been performed on all proposed throshold values;
(a9) checking whether the interpolation has been performed on all pixels of the binary image, if the comparison of the step a6 was performed on all the proposed threshold values, and returning to the step a3 if the interpolation has not been performed on all pixels of the binary image; and
(al0) setting the proposed threshold value providing the highest frequency of hits as the threshold value of the corresponding interpolation value, if the interpolation is performed on all pixels of the binary image.
15

6. An interpolation method for a binary image, substantially as herein described, particularly with reference to and as illustrated in the accompanying drawings.
An improved interpolation method in which a threshold value used for determining a pixel value of a pixel generated by interpolation according to a context (state value of adjacent pixels). In the interpolation method, the ambiguity between the interpolation value and the threshold value is removed by using the context (state value of the reference pixels around the interpolated pixel), thereby reducing the blocking and smoothing phenomena in the restored binary image.

Documents:

02475-cal-1997-abstract.pdf

02475-cal-1997-claims.pdf

02475-cal-1997-correspondence.pdf

02475-cal-1997-description(complete).pdf

02475-cal-1997-drawings.pdf

02475-cal-1997-form-1.pdf

02475-cal-1997-form-2.pdf

02475-cal-1997-form-3.pdf

02475-cal-1997-form-5.pdf

02475-cal-1997-letters patent.pdf

02475-cal-1997-p.a.pdf

02475-cal-1997-priority document others.pdf

02475-cal-1997-priority document.pdf

2475-CAL-1997-FORM-27-1.1.pdf

2475-CAL-1997-FORM-27.pdf


Patent Number 200505
Indian Patent Application Number 2475/CAL/1997
PG Journal Number 07/2007
Publication Date 16-Feb-2007
Grant Date 16-Feb-2007
Date of Filing 30-Dec-1997
Name of Patentee SAMSUNG ELECTRONICS CO.LTD.
Applicant Address 416, MAETAN-DONG,PALDAL-GU, SUWON-CITY, KYUNGKI-DO,
Inventors:
# Inventor's Name Inventor's Address
1 DAE-SUNG CHO 815,SEOUL,GARDEN APT.,49-5,HWIGYUNG 2-DONG, DONGDAEMOON-GU,SEOUL,
2 JAE-SEOB SHIN 104-802, HYUNDAI APT., 707,SHINDAEBANG-DONG,DONGJAK-GU,SEOUL,
PCT International Classification Number G06T 7/40
PCT International Application Number N/A
PCT International Filing date
PCT Conventions:
# PCT Application Number Date of Convention Priority Country
1 97-21781 1997-07-10 Republic of Korea