You want to turn this into an academic pissing contest, or what? Yes, I know that ImageMagick has a "Normalize" function. What I was talking about was normalizing in a much more mathematical sense. In your case, for instance, normalizing the distances from the average for each channel against the square root of the sum of the squares of the distances. There are probably other normalizations that might be useful too, that's just the first one to come to mind.
Also, you want to use YUV because it gives much more useful information in each channel, and you can weight the channels depending on whether you are more interested in brightness or colour differences.