Color similarity sorting for video forensics search

1. A method of sorting electronic color images of objects, the method comprising:

receiving, at a server from an end-user device, an input representation of a desired object to be located in the color images, the input representation including first data representing a first color of the desired object, the first data being in a first color space,

determining an indication of saturation of the first color;

defining a prescribed range of saturation indications based on the determined indication of saturation of the first color;

comparing the prescribed range of saturation indications to a plurality of candidate saturation indications associated with a plurality of stored candidate images; and

in response to the candidate saturation indication associated with a presently compared one of the stored candidate images not being in the prescribed range of saturation indications, the prescribed range of saturation indications including the indication of saturation of the first color, discarding the presently compared candidate image as possibly including the desired object; or

in response to the candidate saturation indication associated with the presently compared one of the stored candidate images being in the prescribed range of saturation indications, determining that the candidate image possibly includes the desired object.

2. The method of claim 1 further comprising:

converting the first data in the first color space into second data in a second color space different than the first color space, wherein at least one of the first color space or the second color space includes a saturation parameter;

determining a query feature vector including a plurality of parameters associated with color of the input representation, the query feature vector parameters including at least a first parameter from the first color space and at least a first parameter from the second color space;

making a feature vector comparison by comparing the query feature vector to a plurality of candidate feature vectors associated with at least a portion of the stored candidate images, each candidate feature vector including a plurality of parameters associated with color of a respective one of the stored candidate images, the candidate feature vector parameters including at least the first parameter from the first color space and at least the first parameter from the second color space; and

in response to the feature vector comparison, determining that the candidate image possibly includes the desired object.

3. The method of claim 2 wherein the input representation is an image representation comprising a plurality of pixels.

4. The method of claim 3 further comprising:

down-sampling the input representation to a block of down-sampled pixels; and

determining a plurality of spatial frequency values of the block of down-sampled pixels in at least one of the first color space or the second color space,

wherein the query feature vector further includes at least one of the spatial frequency values of the input representation image and each of the candidate feature vectors includes at least one spatial frequency value of the respective candidate image.

5. The method of claim 4 wherein the first color space is relative RGB color space, the second color space is HSV, and the spatial frequency values are in YCbCr color space.

6. The method of claim 1 further comprising discarding the presently compared candidate image in response to a value of the indication of saturation of the first color being greater than a threshold value and the candidate saturation indication of the presently compared candidate image being less than the threshold value.

7. The method of claim 1 wherein a user interface allows the user to choose one or more colors for the input representation of the desired object from a color palette.

8. The method of claim 1 wherein the first color corresponds to a first region of the desired object, the method further comprising receiving, at the server from the end-user device, an input representation of a second color of a second region of the desired object, the second region being distinct from the first region.

9. A system for sorting electronic color images of objects, the system comprising:

a communication interface configured to receive, at a server from an end-user device, an input representation of a desired object, the input representation including first data representing a first color of the desired object, the first data being in a first color space;

and

a processing unit communicatively coupled to the communication interface and configured to:

determine an indication of saturation of the first color of the received input representation;

define a prescribed range of saturation indications based on the determined indication of the saturation of the first color;

compare the prescribed range of saturation indications to a plurality of candidate saturation indications associated with a plurality of candidate images stored in a memory of the processing unit;

in response to the candidate saturation indication associated with a presently compared one of the stored candidate images not being in the prescribed range of saturation indications, discard the presently compared candidate image as possibly including the desired object; and

in response to the candidate saturation indication of the presently compared one of the stored candidate images being in the prescribed range of saturation indications, determine that the candidate image possibly includes the desired object.

10. The system of claim 9 wherein the processing unit is configured to convert first data in the first color space into second data in a second color space different than the first color space, wherein at least one of the first color space or the second color space includes a saturation parameter; and

wherein the first color space or the second color space is relative RGB color space.

11. The system of claim 9 wherein the processing unit is further configured to:

determine a query feature vector including a plurality of parameters associated with color of the input representation, the query feature vector parameters including at least a first parameter from the first color space and at least a first parameter from the second color space;

make a feature vector comparison by comparing the query feature vector to a plurality of candidate feature vectors associated with at least a portion of the stored candidate images, each candidate feature vector including a plurality of candidate feature vector parameters associated with color of a respective one of the stored candidate images, the candidate feature vector parameters including at least the first parameter from the first color space and at least the first parameter from the second color space; and

in response to the feature vector comparison, determine that the candidate image possibly includes the desired object.

12. The system of claim 11 wherein the input representation is an image representation comprising a plurality of pixels.

13. The system of claim 12 wherein the processing unit is further configured to:

down-sample the input representation to a block of down-sampled pixels; and

determine a plurality of spatial frequency values of the block of down-sampled pixels in at least one of the first color space or the second color space,

wherein the query feature vector further includes at least one of the spatial frequency values of the input representation image and each of the candidate feature vectors includes at least one spatial frequency value of the respective candidate image.

14. The system of claim 13 wherein the first color space is relative RGB color space, the second color space is HSV, and the spatial frequency values are in YCbCr color space.

15. The system of claim 9 wherein the processing unit is further configured to discard the presently compared candidate image if a value of the saturation indication of the first color is greater than a threshold value and the candidate saturation indication of the presently compared candidate image is less than the threshold value.

16. The system of claim 9 further comprising the end-user device, the end-user device including a user interface configured to allow the user to choose one or more colors for the input representation of a desired object by selecting one or more colors from a color palette.

17. A non-transitory machine-readable storage medium comprising instructions embodied thereon that, when executed by at least one machine, cause the at least one machine to:

receive at a server from an end-user device, an input representation of a desired object, the input representation including first data representing a first color of the desired object, the first data being in a first color space,

determine an indication of saturation of the first color;

define a prescribed range of saturation indications based on the determined indication of saturation of the first color;

compare the prescribed range of saturation indications to a plurality of candidate saturation indications associated with a plurality of stored candidate images;

in response to the candidate saturation indication associated with a presently compared one of the stored candidate images not being in the prescribed range of saturation indications, discard the presently compared candidate image as possibly including the desired object; and

in response to the candidate saturation indication associated with a presently compared one of the stored candidate images being in the prescribed range of saturation indications, determine that the candidate image possibly includes the desired object.

18. The non-transitory machine-readable storage medium recited in claim 17 wherein the instructions further comprise instructions that cause the at least one machine to:

convert the first data representing at least one color of a desired object from the first color space into second data in a second color space different than the first color space, wherein at least one of the first color space or the second color space includes a saturation parameter;

determine a query feature vector including a plurality of parameters associated with color of the desired object, the query feature vector parameters including at least a first parameter from the first color space and at least a first parameter from the second color space;

make a feature vector comparison of the query feature vector to a plurality of candidate feature vectors associated with at least a portion of the stored candidate images, each candidate feature vector including a plurality of parameters associated with color of a respective one of the stored candidate images, the candidate feature vector parameters including at least the first parameter from the first color space and at least the first parameter from the second color space; and

determine based on the feature vector comparison, that the candidate image possibly includes the desired object.

19. The non-transitory machine-readable storage medium recited in claim 18 wherein the first data is an image representation comprising a plurality of pixels.

20. The non-transitory machine-readable storage medium recited in claim 19 wherein the instructions further comprise instructions that cause the at least one machine to:

down-sample the pixels of the image to a block of down-sampled pixels; and

determine a plurality of spatial frequency values of the block of down-sampled pixels in at least one of the first color space or the second color space,

wherein the query feature vector further includes at least one of the spatial frequency values of the input representation image and each of the candidate feature vectors includes at least one spatial frequency value of the respective candidate image.