Image manipulation for web content

1. A computer-implemented method for document generation comprising:

receiving a selection of a plurality of images;

analyzing colors in the plurality of images to determine a palette;

generating a palette based on the analyzing of the colors in the plurality of images;

analyzing shapes of the plurality of images to determine a composition theme;

estimating a predicted click through rate for each of the plurality of images for a given user based at least in part on history data of the given user; and

generating a page including the plurality of images wherein the page is based on the palette, the predicted click through rate, and the composition theme;

wherein the generating the palette comprises:

grouping pixels in each image of the plurality of images into eight groups;

computing an average color for each of the eight groups for each image of the plurality of images to create an image palette for each image of the plurality of images; and

generating the palette by comparing the image palettes for the plurality of images using minimum cost bipartite-matching.

2. The method of claim 1 wherein the composition theme includes aspect ratios for the plurality of images.

3. The method of claim 2 further comprising cropping one or more of the plurality of images.

4. The method of claim 1 further comprising determining a group of themes from which a user can select.

5. The method of claim 1 further comprising evaluating saliency.

6. The method of claim 5 wherein the saliency includes determining relevancy for one or more of the plurality of images and emphasizing the one or more of the plurality of images that are most relevant to the composition theme for the page.

7. The method of claim 5 wherein the saliency includes determining a key feature from one the plurality of images and emphasizing the key feature on the page.

8. The method of claim 7 further comprising recommending an arrangement of the plurality of images based on the key feature that was determined.

9. The method of claim 7 wherein the evaluating saliency identifies an important section for one or more of the plurality of images.

10. The method of claim 7 wherein the evaluating saliency further comprises performing face detection.

11. The method of claim 7 wherein the evaluating saliency further comprises evaluating features within the plurality of images to determine a salient part of a first image and a non-salient part of a second image, wherein generating the page includes overlaying the salient part of the first image over the non-salient part of the second image such that at least some of the salient part occludes at least some of the non-salient part.

12. The method of claim 1 wherein the palette includes a background color for the page.

13. The method of claim 12 wherein the palette is based on an evaluation of distances in a perceptually linear color space.

14. The method of claim 1 further comprising determining a group of palettes from which a user can select.

15. The method of claim 14 further comprising presenting a subset of the group of palettes to the user.

16. The method of claim 15 further comprising receiving a selection for one of the subset of the group of palettes.

17. The method of claim 1 further comprising inserting an advertisement on the page.

18. The method of claim 1 wherein the page is generated to be optimized for a mobile device.

19. The method of claim 1 wherein the history data of the given user comprises past click-through information by the given user.

20. The method of claim 1 wherein the history data of the given user comprises demographic data about the given user.

21. A computer program product embodied in a non-transitory computer readable medium for image manipulation, the computer program product comprising:

code for receiving a selection of a plurality of images;

code for analyzing colors in the plurality of images to determine a palette;

code for generating a palette based on the analyzing of the colors in the plurality of images;

code for analyzing shapes of the plurality of images to determine a composition theme;

code for estimating a predicted click through rate for each of the plurality of images for a given user based at least in part on history data of the given user; and

code for generating a page including the plurality of images wherein the page is based on the palette, the predicted click through rate, and the composition theme;

wherein the code for generating the palette comprises:

code for grouping pixels in each image of the plurality of images into eight groups;

code for computing an average color for each of the eight groups for each image of the plurality of images to create an image palette for each image of the plurality of images; and

code for generating the palette by comparing the image palettes for the plurality of images using minimum cost bipartite-matching.

22. A computer system for image manipulation comprising:

a memory which stores instructions;

one or more processors coupled to the memory wherein the one or more processors are configured to:

receive a selection of a plurality of images;

analyze colors in the plurality of images to determine a palette;

generate a palette based on the analyzing of the colors in the plurality of images;

analyze shapes of the plurality of images to determine a composition theme;

estimate a predicted click through rate for each of the plurality of images for a given user based at least in part on history data of the given user; and

generate page including the plurality of images wherein the page is based on the palette, the predicted click through rate, and the composition theme;

wherein the generating the palette comprises:

grouping pixels in each image of the plurality of images into eight groups;

computing an average color for each of the eight groups for each image of the plurality of images to create an image palette for each image of the plurality of images; and

generating the palette by comparing the image palettes for the plurality of images using minimum cost bipartite-matching.