rgb2ind equivalent for grayscale images - NOT gray2ind because its not the same, at all.

So I love that rgb2ind finds the best matching colors for the number of colors that you want and that you have the option of no dithering, but I realized that finding the best colors in a NxMx3 matrix is probably a lot more computationally expensive than just doing a grayscale image.

The gray2ind just does an even colormap split based on the number of colors you want, not the image itself, for instance, if I want 3 colors, it just bins it like [0 .5 1] which I find not useful at all in this case.

Having both the no dithering option and the color approximation based on image values are important to me, does anyone have a suggestion?

Answer 1

[Image Analyst]

There is a new imquantize() function. You should probably be using imquantize() for grayscale quantization rather than gray2ind(). I don't see any need for gray2ind().

[Shaun VanWeelden]

That and multithresh were exactly what I was looking for, thanks!

Answer 2

[Walter Roberson]

Work around:

   rgb2ind( YourGrayImage(:,:,[1 1 1]), N )

[Shaun VanWeelden]

Do you think that would have a performance benefit? Wouldn't it still have to go over 3 times as much data compared to just using grayscale image values (2-d) ?

[Walter Roberson]

I didn't say it would be efficient performance: it is a work-around, to provide the functionality of finding the "best" colors.

[Shaun VanWeelden]

haha well thank you anyway, I do appreciate it and I hadn't thought of doing grayImg(:,:,[1 1 1]) before, I usually use repmat, that is a nice trick!