Method and apparatus for segmenting multi-view images into foreground and background based on codebook

1. An apparatus comprising:

a background model generation unit for extracting a codebook from multi-view background images and generating codeword mapping tables operating in conjunction with the codebook; and

a foreground and background segmentation unit for segmenting multi-view images into foreground and background using the codebook and the codeword mapping tables.

2. The apparatus of claim 1 , wherein the background model generation unit includes:

a codebook generation unit for extracting codeword indexes from pixel information of the multi-view background images to generate codebook information; and

a codeword mapping table generation unit for generating the codeword mapping tables to which the codeword indexes are mapped.

3. The apparatus of claim 2 , wherein the codebook generation unit measures a distance between pixel information and an existing codeword, determines the pixel information as a new codeword if the distance is equal to or greater than a predetermined critical value, and determines the pixel information as the existing codeword if the distance is less than the predetermined critical value.

4. The apparatus of claim 2 , wherein the codeword mapping table generation unit generates the codeword mapping tables for respective views.

5. The apparatus of claim 1 , wherein the background model generation unit further includes a codeword rearrangement unit for rearranging corresponding indexes for respective pixels of the codeword mapping tables in descending order of frequency of occurrence in the codebook.

6. The apparatus of claim 1 , wherein the foreground and background segmentation unit includes:

a foreground/background determination unit for measuring, when the multi-view images are received, distances between pixel information of respective pixels in the multi-view images and corresponding codeword information based on the codebook and the codeword mapping tables to determine whether the respective pixels correspond to the foreground or the background; and

a post-processing unit for performing post-processing on the pixels, determined as the foreground or the background, using a morphology operation, to segment the multi-view images into the foreground and the background.

7. The apparatus of claim 6 , wherein the foreground/background determination unit sequentially measures the distances from the respective pixels to codewords positioned at the respective pixels, to determine, if the distance is less than a critical value, a corresponding pixel as the background, and determine, if the distance is equal to or greater than the critical value, a corresponding pixel as the foreground.

8. The apparatus of claim 6 , wherein the morphology operation is executed by:

performing an erosion operation which performs binary process on a whole input image to generate a binary image and then reduces a size of the foreground relative to the background in the binary image;

performing an expansion operation which uniformly expands the foreground relative to the background; and

processing a boundary line of each region using a difference between two images output from the erosion and expansion operations.

9. The apparatus of claim 1 , wherein the background model generation unit, when it is based on a color model of a specific image, is set with codewords modeling background color information of the multi-view images, the codebook representing 3-D position information in color space, and the codeword mapping tables which are composed of a codeword number and a codeword index at each position.

10. The apparatus of claim 1 , wherein performance of the segmentation into the foreground and the background is determined based on an error rate obtained by dividing a sum of the number of pixels, corresponding to the background but having been erroneously determined as the foreground, and the number of pixels, corresponding to the foreground but having been erroneously determined as the background, by the number of pixels constituting the foreground.

11. A method comprising:

generating a background model by extracting a codebook from multi-view background images and generating codeword mapping tables operating in conjunction with the codebook; and

segmenting multi-view images into foreground and background using the codebook and the codeword mapping tables.

12. The method of claim 11 , wherein said generating a background model includes:

extracting codeword indexes from pixel information of the multi-view background images to generate codebook information; and

generating the codeword mapping tables to which the codeword indexes are mapped.

13. The method of claim 12 , wherein said segmenting multi-view images includes:

measuring a distance between pixel information and an existing codeword in the codebook;

determining the pixel information, when the distance is equal to or greater than a predetermined critical value, as a new codeword to add the new codeword to the codebook; and

determining the pixel information, when the distance is less than the predetermined critical value, as the existing codeword.

14. The method of claim 12 , wherein the codeword mapping tables are generated for respective views.

15. The method of claim 11 , wherein said generating a background model includes rearranging codeword indexes for respective pixels in the codeword mapping tables in order of occurrence frequency of the codeword indexes in the codebook.

16. The method of claim 11 , wherein said segmenting multi-view images into foreground and background includes:

receiving the multi-view images;

measuring distances between pixel information of respective pixels in the multi-view images and corresponding codeword information based on the codebook and the codeword mapping tables to determine whether the respective pixels correspond to the foreground or the background; and

performing post-processing on the pixels, determined as the foreground or the background, using a morphology operation, to segment the multi-view images into the foreground and the background.

17. The method of claim 16 , wherein said determining whether the respective pixels correspond to the foreground or the background includes:

sequentially measuring the distances from the respective pixels to codewords positioned at the respective pixels;

determining, if the distance is less than a critical value, a corresponding pixel as the background; and

determining, if the distance is equal to or greater than the critical value, a corresponding pixel as the foreground.

18. The method of claim 16 , wherein the morphology operation is executed by:

performing an erosion operation which performs binary process on a whole input image to generate a binary image and then reduces a size of the foreground relative to the background in the binary image;

performing an expansion operation which uniformly expands the foreground relative to the background; and

processing a boundary line of each region using a difference between two images output from the erosion and expansion operations.

19. The method of claim 11 , wherein the background model, when it is based on a color model of a specific image, is set with codewords modeling background color information of the multi-view images, the codebook representing 3-D position information in color space, and the codeword mapping tables which is composed of a codeword number and a codeword index at each position.

20. The method of claim 11 , wherein performance of the segmentation into foreground and background is determined based on an error rate obtained by dividing a sum of the number of pixels, corresponding to the background but having been erroneously determined as the foreground, and the number of pixels, corresponding to the foreground but having been erroneously determined as the background, by the number of pixels constituting the foreground.