Context aware color reduction

1. A method for color reduction, comprising:

segmenting an input image into a plurality of regions including a foreground region;

identifying a class of an object in the foreground region;

assigning a weight to each region, wherein the weight for the foreground region is based on the class of the object;

identifying one or more colors for each of the regions;

selecting a color palette based on the one or more colors for each of the regions and the corresponding weight for each of the regions; and

performing a color reduction on the input image using the selected color palette to produce a color reduced image.

2. The method of claim 1 , further comprising:

identifying a background region of the input image;

identifying one or more objects in the foreground region, wherein the input image is segmented based on the one or more objects; and

classifying the one or more objects, wherein the weight for each of the regions is based at least in part on the classification.

3. The method of claim 2 , wherein:

the foreground region and the background region are identified using a foreground-background separation algorithm.

4. The method of claim 2 , wherein:

the one or more objects are identified using an object classification neural network.

5. The method of claim 1 , further comprising:

generating an image mask for each of the regions.

6. The method of claim 1 , further comprising:

identifying one or more characteristics for each of the regions, wherein the one or more characteristics include relevance, prominence, focus, position, or any combination thereof;

normalizing the one or more characteristics; and

generating the weight for each of the regions by applying a linear regression model to the one or more characteristics.

7. The method of claim 1 , further comprising:

generating a heatmap comprising a visual prominence value for each pixel in the input image; and

identifying a weighted color contribution for each of a plurality of candidate colors in each of the regions based on a pixel count of colors weighted by the visual prominence value, wherein the one or more colors for each of the regions are identified based on the color contribution.

8. The method of claim 7 , wherein:

the one or more colors for each of the regions are identified based on whether the weighted color contribution for each of the candidate colors is above a threshold value.

9. The method of claim 1 , further comprising:

multiplying each of the one or more colors for each of the regions by the corresponding weight assigned to each of the regions to produce a weighted list of colors for each of the regions; and

merging the weighted list of colors for each of the regions to produce a combined weighted list of colors, wherein the color palette is selected based on the combined weighted list of colors.

10. The method of claim 1 , further comprising:

identifying a perceptual contribution of the one or more colors for each of the regions, wherein the color palette is selected based at least in part on the perceptual contribution of the one or more colors for each of the regions.

11. The method of claim 1 , wherein:

the color reduction is performed based on a single-click input from a user.

12. A method for color reduction, comprising:

identifying one or more objects in an input image;

segmenting the input image into a plurality of regions including a foreground region;

identifying a focus characteristic based on a clarity or contrast of an image segment in the input image;

identifying a weight for each pixel of the input image based on the focus characteristic;

calculating a weighted color contribution for each of a plurality of colors based on the identified weights;

selecting a color palette based on the weighted color contribution for each of the plurality of colors; and

performing a color reduction on the input image based on the selected color palette.

13. The method of claim 12 , further comprising:

identifying the foreground region and a background region of the input image;

identifying the one or more objects in the foreground region, wherein the input image is segmented based on the one or more objects; and

classifying the one or more objects.

14. The method of claim 13 , wherein:

the foreground region and the background region are identified using a foreground-background separation algorithm.

15. The method of claim 12 , further comprising:

generating an image mask.

16. The method of claim 12 , further comprising:

identifying one or more characteristics for each of the objects, wherein the one or more characteristics include relevance, prominence, focus, position, or any combination thereof;

normalizing the one or more characteristics; and

applying a linear regression model to the one or more characteristics.

17. The method of claim 12 , further comprising:

generating a heatmap comprising a visual prominence value for each pixel in the input image; and

identifying the weighted color contribution for each of a plurality of colors based on a pixel count of colors weighted by the visual prominence value.

18. The method of claim 12 , further comprising:

multiplying one or more colors by a corresponding weight; and

merging a weighted list of colors for each of the objects.

19. The method of claim 12 , further comprising:

identifying a perceptual contribution of one or more colors for each of the objects.

20. A method for color reduction, comprising:

identifying a foreground and a background of an input image;

identifying one or more objects in the foreground;

classifying the one or more objects;

segmenting the input image into a plurality of regions based on the one or more objects;

identifying one or more object characteristics for each of the regions based on the classification;

normalizing the one or more characteristics;

generating a weight for each of the regions by applying a linear regression model to the one or more characteristics;

generating a heatmap corresponding to a visual prominence of pixels in each of the regions;

identifying a weighted color contribution for each of a plurality of candidate colors in each of the regions based on a pixel count of colors weighted by the visual prominence;

identifying one or more colors for each of the regions based on the color contribution;

multiplying each of the one or more colors for each of the regions by the corresponding weight for each of the regions to produce a weighted list of colors for each of the regions;

merging the weighted list of colors for each of the regions to produce a combined weighted list of colors;

selecting a color palette based on the combined weighted list; and

performing a color reduction on the input image using the selected color palette.