I'm trying to design a histogram kinda function

I have a logo and I would like to retrieve the "most used colors", but filter similar colors.

My idea is like the Kuler from Adobe, where you get 3-5 colors that will look nice together. They are often quite a bit apart in the "colorwheel".

So, I receive a image and I would like to see if there is some kinda way to figure out what primary colors it have and then af开发者_开发百科ter this result try to figure out what colors would look "Kuler" combined with it.

Question:

How do I make a historygram function that will "group similar colors" ??? To build the histogram I guess I would just make an array of "found colors" and then ++ each time found while scanning the image pixel by pixel.

But instead of RGB(xx,yy,zz) in 256 tones, I think this should be roughly "lowered" to eg. 16 values or perhaps less?

Has anyone done similar or do you have any ideas of how to get started?

I would love to see somekinda example in C# as thats my preferede language for this task.

---

Tools such as Kuler use color models to determine what colors are 'similar', what the 'illumination' of a color is, how 'bright' a color is or what its counter color is.

So you might want to read about color models on wikipedia

Secondly see this post