Multi-resolution color-based image search

1. In a digital medium environment for color-based image searching, a method implemented by a processing device, the method comprising:

determining a color histogram of a query image;

determining a multi-resolution color histogram of the query image based in part on the color histogram, the multi-resolution color histogram representing color distribution of the query image;

generating a color vector of the query image by taking square roots of color designations in the multi-resolution color histogram, the generating the color vector including padding a result of the taking the square roots with one or more zeros and compressing the result of the taking the square roots; and

selecting one or more images from searched images based on distance measures between the color vector of the query image and respective color vectors of the searched images, the selected one or more images having only colors that are included in the query image.

2. The method of claim 1 , further comprising:

generating the color vector of the query image by applying a mapping to the multi-resolution color histogram.

3. The method of claim 1 , wherein:

the determining the color histogram as a first three-dimensional histogram in a LAB color space; and

the determining the multi-resolution color histogram as a second three-dimensional histogram in the LAB color space.

4. The method of claim 3 , wherein:

the color histogram includes different respective resolutions in each dimensions of the LAB color space; and

the multi-resolution color histogram includes a same resolution in each of the dimensions of the LAB color space.

5. The method of claim 1 , wherein the generating the color vector includes the compressing the result of the taking the square roots to reduce a length of the result.

6. The method of claim 5 , wherein the compressing includes:

splitting the result of the taking the square roots into multiple vectors in a subspace;

dividing the subspace into clusters;

denoting each vector of the multiple vectors with a respective identifier of one of the clusters; and

concatenating the respective identifier for each vector to generate the color vector.

7. The method of claim 5 , wherein the generating the color vector includes the padding the result of the taking the square roots with the one or more zeros prior to the compressing.

8. The method of claim 1 , wherein the selecting the one or more images includes determining the distance measures as L2-norms between the color vector of the query image and the respective color vectors of the searched images.

9. An image search system for color-based image searching implemented by a processing device in a digital medium environment, the image search system comprising:

a color histogram module to determine a color histogram of a query image, and generate a multi-resolution color histogram of the query image based at least in part on the color histogram;

a matching module to:

determine distance measures between a color vector of the query image and respective color vectors of searched images; and

select one or more images from the searched images having only colors that are included in the query image and the one or more images including no colors that are not included in the query image as determined based on the distance measures that indicate color similarity of the one or more images to the query image.

10. The image search system of claim 9 , wherein the color histogram module is configured to generate the color vector of the query image by applying a mapping to the multi-resolution color histogram.

11. The image search system of claim 10 , further comprising an encoder module to compress the color vector to form a compressed color vector that has a length less than a length of the color vector.

12. The image search system of claim 9 , wherein the color histogram module is configured to determine the color histogram and the multi-resolution color histogram as respective three-dimensional histograms in a LAB color space.

13. The image search system of claim 12 , wherein the color histogram includes different respective resolutions in each dimensions of the LAB color space, and set the multi-resolution color histogram includes a same resolution in each of the dimensions of the LAB color space.

14. The image search system of claim 9 , further comprising an encoder module to:

pad the color vector of the query vector with one or more zeros;

split the padded color vector into multiple vectors in a subspace that has a dimension less than a dimension of the padded color vector;

divide the subspace into clusters;

denote each vector of the multiple vectors with a respective identifier of one of the clusters; and

concatenate the respective identifier for each vector to form a compressed color vector that has a length less than a length of the color vector.

15. The image search system of claim 14 , wherein the matching module is configured to generate the distance measures as L2-norms between the compressed color vector of the query image and the respective color vectors of the searched images.

16. One or more non-transitory computer-readable media storing a plurality of executable instructions, which when executed by a processing device, cause the processing device to perform operations comprising:

determining a multi-resolution color histogram of a query image, the multi-resolution color histogram representing color distribution of the query image;

generating a color vector of the query image by taking square roots of color designations in the multi-resolution color histogram of the query image, and padding a result of the taking the square roots with one or more zeros;

generating a compressed color vector of the query image by compressing the color vector to reduce a length of the color vector;

determining distance measures as L2-norms between the compressed color vector of the query image and respective color vectors of searched images; and

selecting one or more images from the searched images based on the distance measures that indicate color similarity of the one or more images to the query image.

17. The one or more non-transitory computer-readable media of claim 16 , wherein the padding the result of the taking the square roots with the one or more zeros is prior to the compressing the color vector.

18. The one or more non-transitory computer-readable media of claim 17 , wherein the selected one or more images have only colors that are included in the query image.

19. The one or more non-transitory computer-readable media of claim 16 , wherein the selected one or more images include no colors that are not included in the query image.

20. The one or more non-transitory computer-readable media of claim 16 , wherein the determining the multi-resolution color histogram as a three-dimensional histogram in LAB color space.