12:23 PM 3/28/02 So we're on technique 4. In this technique, further grouping will occur. Basically the algorithm goes as follows: Use 37 mask Compare every color against each other inside the mask.(Can be sped up *greatly* with optimization later) And its +/- 20 color on GREEN, BLUE, OR RED... Then its reclassified as the first color. The image is processed onto itself, such that future masks use the previous possible colors, and the masks overlap by 2 extra pixels this time instead of being spaced out 5. (leads to a possible recoding of technique #1 that uses overlap)

LEFT: My starting image will be the result of technique #3
RIGHT: Processed with technique #4

Obviously I made an error with my coding somewhere. Looking back through it. I think I gave too much of a range on what is acceptable to group... Narrowing the range some...
LEFT: My starting image will be the result of technique #3
RIGHT: Processed with corrected range on technique #4

Looks better, but still not what I had in mind...
I need to figure this out...
Tryin with the original image instead of after technique #1 and #3 were applied...
Left: My original image
Right: After technique #4

Ok... Whats wrong here... There are obvious vertical streaks where the image blends... I think this is due to the loop going up and down before left and right... Changing the code to go left to right then up and down to see if I can move the lines to become horizontal.... Nope that didn't do it, but I found I can reverse a loop to get rid of the downward streaking tendency. Left: My original image
Right: After technique #4 revised

LEFT: My starting image will be the result of technique #3
Right: After technique #4 revised

Obviously, its doing something... But what its doing really isn't anything I can use... I think the problem comes with the fact that its trying to group colors locally. I played with technique 4 some more.
LEFT: Increased range of colors RIGHT: Then I increased the overlapping of the mask
I don't think technique 4 is a good idea.