How to achieve similar color distribution with fewer pixels?

Question

I've got a set of 200,000 pixels which are mostly Gaussian distributed. Is it possible to efficiently reduce the set of pixels to a set of let's say 1000 pixels which are similarly distributed to the entire set?

Answer 1

The simplest approach is just to randomly select 1,000 pixels from the original 200,000 using a simple random sample. The random sample will have a statistically similar distribution to the original set.

If you want to take a more sophisticated approach that is less subject to random variation and doesn't contain duplicate colors, use a Color Quantization algorithm such as Median Cut to quantize from 200K down to 1000 colors.