Unique Colors

There are about 16 million colors in the 24 bit RGB color scheme. Our eyes can only see between 35,000 and 55,000 unique colors. This means that you will not be able see small changes in color. If there are a large number of pixels of a single color, then it will require a lot of change from the original color and become noticeable.

Galleries

I searched Google extensively for a program that already does this but I failed to find anything. I was curious to see if (A) it could be done and (B) how well it could be done.

I learned that it is possible to convert an image so that every color is unique and yet have it look exactly like the original but only for images that had a high percentage of unique colors to start. If the image had a lot of one color for example black or white, then the deviation would be so great as to introduce noticeable noise. Blurred areas also proved to be a problem since when you blue an image, you reduce the number of unique colors.

Examples of Improving Algorithm

[ Grinder Girl ]   [ Iron Man ]   [ Dominoes ]

I learned that if I purposely add noise to the image before converting it to all unique colors, that it could eliminate the jpeg artifacts and smooth out the blotchy patches so the resulting picture would look like a grainy version of the original. These are some examples of how my algorithm is improving as I learn from my experiments.

The algorithm is described on my developer blog.