Multi-color dropout for scanned document
View Patent ↗A method for removing unwanted form color content from a scanned document by segmenting the image into two or more tiles ( 14 ). Each tile ( 14 ) is classified into at least a first and second set according to its combination of background and color content. A background color is identified from the first set of tiles ( 14 ). At least one form color is identified from the second set of tiles. A transform is applied that shifts form color image data values toward background color data values.
1. A method for removing unwanted form color content from scanned image data for a document, the method comprising:
a) obtaining the scanned document image data in a color data format;
b) segmenting the image data into two or more tiles;
c) characterizing each tile by its color content with the steps of:
(i) generating a red histogram for red color values within the tile;
(ii) generating a green histogram for green color values within the tile;
(iii) generating a blue histogram for blue color values within the tile;
(iv) determining, according to the red, green, and blue histogram distributions, whether the tile contains predominantly neutral background content, or predominantly form color content, or a combination of background and color content;
d) identifying a background color according to tiles having predominantly neutral background content;
e) identifying at least one form color according to tiles having predominantly form color content; and
f) selectively transforming pixels having the at least one form color toward the background color.
2. The method of claim 1 wherein selectively transforming comprises applying coefficients that shift form color values toward background color values.
3. The method of claim 1 wherein identifying the background color comprises:
a) transforming RGB color data to HSV color data; and
b) finding the most common hue color data value for pixels having a hue color data value within about +/−10 degrees of the background color.
4. The method of claim 1 wherein determining, according to the red, green, and blue histogram distributions, whether the tile contains predominantly neutral background content, or predominantly form color content, or a combination of background and color content comprises identifying one or more peak values in the red, green, or blue histogram distributions.
5. The method of claim 1 wherein identifying at least one form color comprises converting an RGB coordinate value to an RSV coordinate value.
6. The method of claim 1 wherein characterizing each tile further comprises reducing the resolution of the image data within the tile.
7. The method of claim 1 wherein the color format is red, green, blue (RGB) format.
8. The method of claim 1 wherein the tiles are substantially non-overlapping.
9. The method of claim 1 wherein the same steps are executed for processing each document in a succession of documents.
10. The method of claim 1 wherein selectively transforming pixels comprises generating transform coefficients based on maximum and minimum RGB data values from the identified form color and applying the transform coefficients to each pixel's maximum and minimum RGB data values.