Content based image retrieval
A method and non-transitory computer readable medium for content based image retrieval. The method includes selecting a query image, segmenting the selected query image by applying a segmentation technique, extracting features from the segmented query image by determining at least two feature descriptors, including color feature descriptors and texture feature descriptors, and determining a similarity of the query image to a plurality of images included in a database using the determined at least two feature descriptors of the segmented query image, features being extracted from each of the plurality of images included in the database by determining the at least two feature descriptors, the color feature descriptors and the texture feature descriptors including a simultaneous combination of different color spaces, and global and local statistical measurements being carried out on the simultaneous combination of the different color spaces.
1. A method for content based image retrieval, the method comprising:
selecting a query image;
segmenting the selected image by applying a segmentation technique;
extracting features from the segmented image by determining at least two feature descriptors, including a color feature descriptor and a texture feature descriptor; and determining, using the determined at least two feature descriptors of the segmented query image, a similarity of the query image to a plurality of images included in a database, which also include features extracted and computed by said at least two descriptors,
the method being characterized in that the determined color feature descriptor and texture feature descriptor comprise global and local statistical measurements of at least the mean and variance of the numerical values for each of three color channels of each one of at least three different color spaces.
2. The method according to claim 1 , wherein the color feature descriptor and texture feature descriptor are further combined with at least a shape feature descriptor and an orientation feature descriptor.
3. The method according to claim 1 , wherein the determined color feature descriptor and texture feature descriptor may comprise only global statistical measurements of at least the mean and variance of the numerical values for each of three color channels of each one of at least three different color spaces.
4. The method according to claim 1 , further comprising, computing a local region of interest, or ROI, of a certain ROI size within the query image in order to distinguish different color and/or texture regions in the query image, wherein the ROI is computed by measuring mean and variance, where the ROI satisfies a distance nearest or farthest to the global statistical measurements.
5. The method according to claim 4 , wherein the ROI size is a percentage of a total area of the query image.
6. The method according to claim 1 , wherein the different color spaces combined simultaneously comprise at least HSV, CieLAB and CieXYZ.
7. The method according to claim 2 , wherein the shape feature descriptor comprises using low-level invariant Hu moments and other non-linear and dimensionless parameters related to a volume of an object, based on areas comprising convex hulls, eccentricities, other derivate computed shapes, and/or ratios thereof related to the object.
8. The method according to claim 7 , further comprising computing:
equivalent perimeter-based ratios and detailed local shapes for inside contours and other high-level variables; and/or
convexity defect areas and related linear ratios for inside and outside borders, convexities and/or eccentricities.
9. The method according to claim 2 , wherein the orientation feature descriptor is determined using at least one angle value of the first component of a principal component analysis, or PCA.
10. The method according to claim 1 , wherein the similarity of the query image to the plurality of images included in the database is determined using space metrics applied within a real-time tailored hyper-cube cluster, the hyper-cube cluster being:
defined by a vector with an m-dimensional radius, of a subset m of the total n-dimensions of the at least two feature descriptors together, with independent and different radius values for each dimension;
based on an accepted distance percentage of a value of a query image position; and
limited by a computation of a maximum speed permitted in order to compute all the necessary 1:1 distances within the hyper-cube cluster, wherein the hyper-cube cluster is centered on the space position of the query image.
11. The method according to claim 10 , wherein the m-dimensional radius vector is computed for each dimension using at least the query image position, changing mathematically and geometrically from a multidimensional space to a one-dimensional space.
12. The method according to claim 1 , wherein the at least two feature descriptors are optimized by applying transform techniques and filtering methods based on statistical classification methods applied to a multi-variant logistical regression analysis.
13. A non-transitory computer readable medium storing a program causing a computer to execute a method for content based image retrieval, the method comprising:
selecting a query image;
segmenting the selected query image by applying a segmentation technique;
extracting features from the segmented query image by determining at least two feature descriptors, including a color feature descriptor and a texture feature descriptor; and
determining, using the determined at least two feature descriptors of the segmented query image, a similarity of the query image to a plurality of images included in a database, which also include features extracted and computed by said at least two descriptors,
wherein the determined color feature descriptor and texture feature descriptor comprise global and local statistical measurements of at least the mean and variance of the numerical values for each of three color channels of each one of at least three different color spaces.
14. The non-transitory computer readable medium according to claim 13 , wherein the color feature descriptor and texture feature descriptor are further combined with at least a shape feature descriptor and an orientation feature descriptor.
15. The non-transitory computer readable medium according to claim 13 , wherein the determined color feature descriptor and texture feature descriptor may comprise only global statistical measurements of at least the mean and variance of the numerical values for each of three color channels of each one of at least three different color spaces.
16. The non-transitory computer readable medium according to claim 13 , further comprising computing a local region of interest, or ROI, of a certain ROI size within the query image in order to distinguish different color and/or texture regions in the query image, wherein the ROI is computed by measuring mean and variance, where the ROI satisfies a distance nearest or farthest to the global statistical measurements.
17. The non-transitory computer readable medium according to claim 16 , wherein the ROI size is a percentage of a total area of the query image.
18. The non-transitory computer readable medium according to claim 13 , wherein the different color spaces combined simultaneously comprise at least HSV, CieLAB, and CieXYZ.
19. The non-transitory computer readable medium according to claim 14 , wherein the shape feature descriptor comprises using low-level invariant Hu moments and other non-linear and dimensionless parameters related to a volume of object, based on areas comprising convex hulls, eccentricities, other derivate computed shapes, and/or ratios thereof related to the object.
20. The non-transitory computer readable medium according to claim 19 , further comprising computing:
equivalent perimeter-based ratios and detailed local shapes for inside contours and other high-level variables; and/or
convexity defect areas and related linear ratios for inside and outside borders, convexities, and/or eccentricities.
21. The non-transitory computer readable medium according to claim 14 , wherein the orientation feature descriptor is determined using at least one angle value of a first component of a principal component analysis, or PCA.
22. The non-transitory computer readable medium according to claim 13 , wherein the similarity of the query image to the plurality of images included in the database is determined using space metrics applied within a real-time tailored hyper-cube cluster, the hyper-cube cluster being:
defined by a vector with an m-dimensional radius, of a subset m of the total n-dimensions of the at least two feature descriptors together, with independent and different radius values for each dimension;
based on an accepted distance percentage of a value of a query image position; and
limited by a computation of a maximum speed permitted in order to compute all necessary 1:1 distances within the hyper-cube cluster, wherein the hyper-cube cluster is centered on the query image position.
23. The non-transitory computer readable medium according to claim 22 , wherein the m-dimensional radius vector is computed for each dimension using at least the query image position, changing mathematically and geometrically from a multidimensional space to a one-dimensional space.
24. The non-transitory computer readable medium according to claim 13 , wherein the at least two feature descriptors are optimized by applying transform techniques and filtering methods based on statistical classification methods applied to a multi-variant logistical regression analysis.