Color Histogram weighted by non-isotropic Gaussian Kernel

I have a patch to extract color histogram, meanwhile I want to incorporate the spatial weights into the color histogram. The spatial weight is decided by a non-isotropic Gaussian kernel, how could I add the weight for corresponding pixels when computing color histogram?

Answer 1

[Image Analyst]

How about multiplying something by the weight and adding it to whatever you're going to add it to. That's about all I can say given this vague description of an algorithm and the lack of any code, pseudocode, or equation.

[Rui Zhao]

Just solved it. After dividing the image into small blocks, and computing their color histogram, I can weight each color histogram by the total energy of the kernel within that block. Then sum all histograms up or just concatenate them, I can get the weighted color histogram.

Sorry for the vague description because I didn't have the clear idea then.