Methods and systems for refining text color in a digital image

1. A method for determining a refinement color value for a pixel of interest in a digital image, said method comprising:

in a computing system comprising a computing device, from a digital image comprising a first plurality of pixels, wherein each pixel in said first plurality of pixels comprises an associated color value, selecting a pixel of interest from said first plurality of pixels, said pixel of interest comprising a pixel-of-interest color value;

in said computing system, separating said first plurality of pixels into a second plurality of pixels and a third plurality of pixels based on said pixel-of-interest color value;

in said computing system, calculating a refinement color value based on said second plurality of pixels when said second plurality of pixels meets a quantity threshold;

in said computing system, calculating said refinement color value based on said third plurality of pixels when said second plurality of pixels does not meet said quantity threshold; and

in said computing system, assigning said refinement color value to said pixel of interest.

2. A method according to claim 1 , wherein said pixel of interest is the center pixel in an examination window comprising said first plurality of pixels.

3. A method according to claim 1 , wherein said first plurality of pixels are text pixels.

4. A method according to claim 3 further comprising assigning said refinement color value to said first plurality of pixels.

5. A method according to claim 1 , wherein said second plurality of pixels comprises pixels from said first plurality of pixels with associated color values substantially similar to said pixel-of-interest color value.

6. A method according to claim 5 further comprising assigning said refinement color value to said second plurality of pixels.

7. A method according to claim 1 , wherein said calculating a refinement color value based on said second plurality of pixels comprises a computation selected from the group consisting of a mean of the associated color values of the pixels in said second plurality of pixels, a median of the associated color values of the pixels in said second plurality of pixels, a weighted average of the associated color values of the pixels in said second plurality of pixels and a trimmed-mean of the associated color values of the pixels in said second plurality of pixels.

8. A method according to claim 1 , wherein said calculating a refinement color value based on said third plurality of pixels comprises a computation selected from the group consisting of a mean of the associated color values of the pixels in said third plurality of pixels, a median of the associated color values of the pixels in said third plurality of pixels, a weighted average of the associated color values of the pixels in said second plurality of pixels and a trimmed-mean of the associated color values of the pixels in said third plurality of pixels.

9. A method according to claim 1 , wherein said separating said first plurality of pixels into said second plurality of pixels and said third plurality of pixels based on said pixel-of-interest color value comprises:

calculating a similarity measure between said pixel-of-interest color value and said associated color value of a second pixel from said first plurality of pixels; and

associating said second pixel with said second plurality of pixels when said similarity measure meets a similarity criterion.

10. A method according to claim 9 , wherein said similarity measure is a distance measure selected from the group consisting of L 1 norm, L 2 norm, Euclidean distance, city-block distance, weighted city-block distance and weighted Euclidean distance.

11. A system, embodied on a non-transitory computer-readable medium, for determining a refinement color value for a pixel of interest in a digital image, said system comprising:

a pixel-of-interest selector for, from a digital image comprising a first plurality of pixels, wherein each pixel in said first plurality of pixels comprises an associated color value, selecting a pixel of interest from said first plurality of pixels, said pixel of interest comprising a pixel-of-interest color value;

a classifier for separating said first plurality of pixels into a second plurality of pixels and a third plurality of pixels based on said pixel-of-interest color value;

a refinement color calculator for:

calculating a refinement color value based on said second plurality of pixels when said second plurality of pixels meets a quantity threshold; and

calculating said refinement color value based on said third plurality of pixels when said second plurality of pixels does not meet said quantity threshold; and

color-value assigner for assigning said refinement color value to said pixel of interest.

12. A method for determining a color value for a pixel of interest in a digital image, said method comprising:

in a computing system comprising a computing device, from a digital image comprising a first plurality of pixels, wherein each pixel in said first plurality of pixels comprises an associated color value, selecting a pixel of interest from said first plurality of pixels, said pixel of interest comprising a pixel-of-interest color value;

in said computing system, identifying from said first plurality of pixels a second plurality of pixels, wherein said color values associated with said second plurality of pixels are substantially dissimilar to a background-color estimate;

in said computing system, calculating a refinement color value based on said second plurality of pixels; and

in said computing system, assigning said refinement color value to said pixel of interest.

13. A method according to claim 12 further comprising calculating said background-color estimate.

14. A method according to claim 12 further comprising calculating said refinement color value based on said second plurality of pixels when said second plurality of pixels meets a quantity threshold.

15. A method according to claim 12 further comprising removing said pixel of interest from an associated text map when said pixel of interest color value is substantially similar to said background-color estimate.

16. A method according to claim 12 , wherein said pixel of interest is the center pixel in an examination window comprising said first plurality of pixels.

17. A method according to claim 12 , wherein said first plurality of pixels are text pixels.

18. A method according to claim 17 further comprising assigning said refinement color value to said first plurality of pixels.

19. A method according to claim 12 further comprising assigning said refinement color value to said second plurality of pixels.

20. A method according to claim 12 , wherein said calculating a refinement color value based on said second plurality of pixels comprises a computation selected from the group consisting of a mean of the associated color values of the pixels in said second plurality of pixels, a median of the associated color values of the pixels in said second plurality of pixels, a weighted average of the associated color values of the pixels in said second plurality of pixels and a trimmed-mean of the associated color values of the pixels in said second plurality of pixels.

21. A method according to claim 12 , wherein said refinement color value comprises a maximally dissimilar color value from said associated color values of said second plurality of pixels relative to said background-color estimate.

22. A method according to claim 12 , wherein said identifying from said first plurality of pixels a second plurality of pixels comprises:

calculating a similarity measure between said background-color estimate and said associated color value of a second pixel from said first plurality of pixels; and

associating said second pixel with said second plurality of pixels when said distance measure meets a dissimilarity criterion.

23. A method according to claim 22 , wherein said similarity measure is a distance measure selected from the group consisting of L 1 norm, L 2 norm, Euclidean distance, city-block distance, weighted city-block distance and weighted Euclidean distance.