[Python] How to compare noise between two RGB images using KL divergence or other similarity...

I have two RGB images that represent noise, and their shape is [H, W, 3] (height, width, 3 color channels). I want to compare how similar the noise distributions are between the two images.

My questions are:

Can I use KL divergence (Kullback–Leibler divergence) to compare the noise patterns of these two RGB images?

Since these are not explicitly modeled distributions, what is the proper way to estimate their probability distributions before computing KL divergence? Are there alternative metrics better suited for comparing noise similarity between images, especially when noise is high-dimensional and not normally distributed?



Continue reading...