Method of image segmentation for object-based image retrieval

Claim 3

The method as claimed in claim 1 , wherein the extraction of color regions from the multi-dimensional image further comprises: applying a median filter to the multi-dimensional image to reduce pixel spike noise; quantizing the color values of the multi-dimensional image; and grouping adjacent pixels into color regions by grouping pixels together based on quantization values.

Claim 4

The method as recited in claim 2 , wherein the size of the median filter is 2×2.

Claim 5

The method as recited in claim 2 , wherein the grouping of pixels into color regions further comprises determining if a color region is small, and if that determination is true, merging the small color region with an adjacent color region closest in color value.

Claim 6

The method as recited in claim 1 , wherein the calculation of complexities between color region pairs further comprises: extracting the boundary lines from the multi-dimensional image; counting the number of boundary pixels of the color region and the number of pixels in the color region immediately adjacent to the color region; and calculating a ratio based on the number of boundary pixels of the color region and the number of pixels in the color region immediately adjacent to the color region to determine the complexity between the color regions.

Claim 7

The method as recited in claim 1 , wherein the merging color regions using hierarchical cluster analysis further comprises: calculating the features of each color region; calculate threshold values based on the feature calculation of the color regions; merge color regions based on the feature distance using a nearest neighbor analysis; calculate threshold values of the merged color regions; and merge color regions based on the feature distance using a furthest neighbor analysis.

Claim 8

The method as recited in claim 6 , wherein the calculation of the features of each color region comprises calculating location, color values and texture values for each of the color regions.

Claim 9

The method as recited in claim 7 , wherein the calculation of the location of the color regions comprises defining the center of gravity of the evaluated color regions and the size of the multi-dimensional image.

Claim 10

The method as recited in claim 7 , wherein the color distance of the color regions is calculated using the Luv color space.

Claim 11

The method as recited in claim 7 , wherein the calculation of the texture of the color regions comprises calculating average and standard deviations of the texture coefficients.

Claim 12

The method as recited in claim 10 , wherein the calculation of the average and standard deviations use the coefficients of the pixels at a predefined depth from the region boundary.

Claim 13

The method as recited in claim 10 , wherein the calculation of the texture distance of the color regions is calculated using the coefficient of the wavelet.

Claim 14

The method as recited in claim 12 , wherein the calculation of the texture of the color regions comprises selecting the pixels at a predefined depth from the region boundary and calculate the average and standard deviation of the wavelet coefficient of the selected pixels.

Claim 15

The method as recited in claim 1 , wherein the merging color regions together using boundary analysis further comprises merging color regions together using the complexity of adjacent color regions.

Claim 16

The method as recited in claim 1 , wherein the merging color regions together using boundary analysis further comprises merging color regions together using the density of the color region boundary pixels.

Claim 17

The method as recited in claim 1 , wherein the merging color regions together using boundary analysis further comprises merging color regions together using the description length of the color description and the boundary description.

Claim 18

The method as recited in claim 1 , wherein the merging color regions together using boundary analysis further comprises: merging color regions together using the complexity of adjacent color regions; merging color regions together using the density of the color region boundary pixels; and merging color regions together using the description length of the color description and the boundary description.

Claim 19

The method as recited in claim 17 , wherein the merging color regions together using boundary analysis further comprises: calculating the color description length and the boundary description length; and merging the color regions together in order to minimize the sum of the color description length and the boundary description length.

Claim 20

The method as recited in claim 17 , wherein the merging color regions together using boundary analysis further comprises: calculating the boundary description reduction ratio and the color description reduction ratio; and merging the color regions together when the sum of the boundary description reduction ratio and the color description reduction ratio is less than a threshold value.

Claim 21

The method as recited in claim 1 , wherein the merging color regions together using boundary analysis further comprises: calculating the location, color values and texture values of the color regions; calculating complexity values for all adjacent color regions; merging the color region based on the region distance and the threshold value; calculate the boundary description reduction ratio and the color description reduction ratio for four adjusting regions; and if the sum of the boundary description reduction ratio and the color description reduction ratio is less than a threshold value, merge the four adjusting regions.

Claim 22

The method as recited in claim 20 , further comprising: after the merge of the four adjusting regions, calculate the boundary description reduction ratio and the color description reduction ratio for three adjusting regions; and if the sum of the boundary description reduction ratio and the color description reduction ratio is less than a threshold value, merge the three adjusting regions.

Claim 23

The method as recited in claim 21 , further comprising: after the merge of the three adjusting regions, calculate the boundary description reduction ratio and the color description reduction ratio for two adjusting regions; and if the sum of the boundary description reduction ratio and the color description reduction ratio is less than a threshold value, merge the two adjusting regions.

Claim 24

The method as recited in claim 22 , further comprising: after the merge of the two adjusting regions, calculate the density ratio for the color regions; merge adjacent color regions if the density ratio is greater than the density threshold; calculate the boundary description reduction ratio and the color description reduction ratio for two adjusting regions; and if the sum of the boundary description reduction ratio and the color description reduction ratio is less than a threshold value, merge the two adjusting regions.

Claim 25

The method as recited in claim 1 , wherein assigning region data to each color region further comprises assigning visual features to each color region.

Claim 26

The method as recited in claim 24 , wherein assigning visual features to each color region further comprises assigning a dominant color descriptor to each color region.

Claim 27

The method as recited in claim 24 , wherein assigning visual features to each color region further comprises assigning a color layout descriptor to each color region.

Claim 28

The method as recited in claim 24 , wherein assigning visual features to each color region further comprises assigning a shape descriptor to each color region.

Claim 29

The method as recited in claim 24 , wherein assigning region data to each color region further comprises calculating an edge histogram descriptor for each color region.

Claim 30

The method as recited in claim 1 , wherein assigning region data to each color region further comprises assigning at least one of a keyword, a tag or an identifier name to each color region.

Claim 31

A computer software product for segmenting a multi-dimensional image, wherein the computer software product comprises: software instructions that enable a computer to perform predetermined operations, and a computer readable medium bearing the software instructions, wherein the predetermined operations comprise: extracting color regions from the multi-dimensional image based on the color values of the pixels; calculating complexities of the boundary between all possible color region pairs, and merging region pairs together if a boundary complexity value of the color region pair is above a predetermined boundary complexity threshold and the color distance between the region pair is less than a predetermined distance threshold value; merging color regions together using hierarchical cluster analysis; merging color regions together using boundary analysis; determining if a color region is small and surrounded by another color region, and if that determination is true, merging the color region into the surrounding color region; and assigning region data to each color region.

Claim 32

A computer software product for segmenting a multi-dimensional image, wherein the computer software product comprises: software instructions that enable a computer to perform predetermined operations, and a computer readable medium bearing the software instructions, wherein the predetermined operations comprise: extracting color regions from the multi-dimensional image based on the color values of the pixels; calculating complexities of the boundary between all possible color region pairs, and merging region pairs together if a boundary complexity value of the color region pair is above a predetermined boundary complexity threshold and the color distance between the region pair is less than a predetermined distance threshold value; merging color regions together using hierarchical cluster analysis; merging color regions together using boundary analysis; determining if a color region is small and surrounded by another color region, and if that determination is true, merging the color region into the surrounding color region; and assigning region data to each color region.

Claim 33

The computer software product as claimed in claim 30 , wherein the extraction of color regions from the multi-dimensional image further comprises: applying a median filter to the multi-dimensional image to reduce pixel spike noise; quantizing the color values of the multi-dimensional image; and grouping adjacent pixels into color regions by grouping pixels together based on quantization values.

Claim 34

The computer software product as claimed in claim 30 , wherein the grouping of pixels into color regions further comprises determining if a color region is small, and if that determination is true, merging the small color region with an adjacent color region closest in color value.

Claim 35

The computer software product as claimed in claim 30 , wherein the calculation of complexities between color region pairs further comprises: extracting the boundary lines from the multi-dimensional image; counting the number of boundary pixels of the color region and the number of pixels in the color region immediately adjacent to the color region; and calculating a ratio based on the number of boundary pixels of the color region and the number of pixels in the color region immediately adjacent to the color region to determine the complexity between the color regions.

Claim 36

The computer software product as claimed in claim 30 , wherein the merging color regions using hierarchical cluster analysis further comprises: calculating the features of each color region; calculate threshold values based on the feature calculation of the color regions; merge color regions based on the region feature distance using a nearest neighbor analysis; calculate threshold values of the merged color regions; and merge color regions based on the region feature distance using a furthest neighbor analysis.

Claim 37

The computer software product as claimed in claim 30 , wherein assigning region data to each color region further comprises assigning at least one visual feature to each color region.

Claim 38

The computer software product as claimed in claim 35 , wherein assigning visual features to each color region further comprises assigning at least one of a dominant color descriptor, a color layout descriptor, a shape descriptor or an edge histogram descriptor for each color region.

Claim 39

The computer software product as claimed in claim 30 , wherein assigning region data to each color region further comprises assigning at least one of a keyword, a tag or an identifier name to each color region.

Claim 40

A computer system for segmenting a multi-dimensional image, wherein the computer system comprises: a processor; and a memory that stores software instructions that are executed by the processor, the software instructions comprising: extracting color regions from the multi-dimensional image based on the color values of the pixels; calculating complexities of the boundary between all possible color region pairs, and merging region pairs together if a boundary complexity value of the color region pair is above a predetermined boundary complexity threshold and the color distance between the region pair is less than a predetermined distance threshold value; merging color regions together using hierarchical cluster analysis; merging color regions together using boundary analysis; determining if a color region is small and surrounded by another color region, and if that determination is true, merging the color region into the surrounding color region; and assigning region data to each color region.

Claim 41

A computer system for segmenting a multi-dimensional image, wherein the computer system comprises: a processor; and a memory that stores software instructions that are executed by the processor, the software instructions comprising: extracting color regions from the multi-dimensional image based on the color values of the pixels; calculating complexities of the boundary between all possible color region pairs, and merging region pairs together if a boundary complexity value of the color region pair is above a predetermined boundary complexity threshold and the color distance between the region pair is less than a predetermined distance threshold value; merging color regions together using hierarchical cluster analysis; merging color regions together using boundary analysis; determining if a color region is small and surrounded by another color region, and if that determination is true, merging the color region into the surrounding color region; and assigning region data to each color region.

Claim 42

The computer software product as claimed in claim 38 , wherein the extraction of color regions from the multi-dimensional image further comprises: applying a median filter to the multi-dimensional image to reduce pixel spike noise; quantizing the color values of the multi-dimensional image; and grouping adjacent pixels into color regions by grouping pixels together based on quantization values.

Claim 43

The computer software product as claimed in claim 38 , wherein the grouping of pixels into color regions further comprises determining if a color region is small, and if that determination is true, merging the small color region with an adjacent color region closest in color value.

Claim 44

The computer software product as claimed in claim 38 , wherein the calculation of complexities between color region pairs further comprises: extracting the boundary lines from the multi-dimensional image; counting the number of boundary pixels of the color region and the number of pixels in the color region immediately adjacent to the color region; and calculating a ratio based on the number of boundary pixels of the color region and the number of pixels in the color region immediately adjacent to the color region to determine the complexity between the color regions.

Claim 45

The computer software product as claimed in claim 38 , wherein the merging color regions using hierarchical cluster analysis further comprises: calculating the features of each color region; calculate threshold values based on the feature calculation of the color regions; merge color regions based on the region feature distance using a nearest neighbor analysis; calculate threshold values of the merged color regions; and merge color regions based on the region feature distance using a furthest neighbor analysis.

Claim 46

The computer software product as claimed in claim 38 , wherein assigning region data to each color region further comprises assigning at least one visual feature to each color region.

Claim 47

The computer software product as claimed in claim 43 , wherein assigning visual features to each color region further comprises assigning at least one of a dominant color descriptor, a color layout descriptor, a shape descriptor or an edge histogram descriptor for each color region.

Claim 48

The computer software product as claimed in claim 38 , wherein assigning region data to each color region further comprises assigning at least one of a keyword, a tag or an identifier name to each color region.