Three patents have been assigned to my former employer. They are in various stages of prosecution. The topics are in the area of image processing algorithms and hardware. It is not necessarily in the interest of the assignee to have patents published before products are in the field.
I can only comment on those with public
documentation.
The CIELAB color space is used to search for colors in a document that would print at the same black and white level. CIE allows me to make an easy to search 3 dimension table to check for the existence of those colors. Textures added to differentiate colors are noise and become obvious in competing systems when you look at something like a color sweep. Other competing systems are implemented as optional settings for the user. That makes them of marginal usefulness for novice users.
In a real system there is a severe limit placed by the scanner. Scanners have significant pixel to pixel and systematic intensity variations. Any implementation of this algorithm would require some tuning to account for scanner variations and just noticeable differences on the output.
A next obvious step is to only histogram pixels based on a segmented input. I.e. look for and only process smooth areas.
I have to thank Ballard and Brown for producing an excellent book on Computer Vision ( Prentice-Hall, 1982.) It contains an algorithm on region growing. The algorithm is simple and easy to implement in hardware.
A xerographic system only produces black and white dots. Those are only produced at some limited resolution. To achieve a pleasing final result the text, continuous tone, and half toned areas of an original must be treated differently. A region growing step can eliminate unpleasant switches between processing modes.
A next step would be to use the windows to set things like white and black points for the region. Knowledge of those characteristics is critical for systems with limited dynamic range.