I'm not sure of your definition of spatial color contrast, but you could construct a delta E image. Convert the image from RGB to LAB with makecform. Then convolve each L, A, and B image with a kernel like [1 1 1; 1 0 1; 1 1 1]/8 to get the average L, A, or B in the 8 neighbors surrounding each pixel. Then calculate deltaL image, deltaA image, and the deltaB image by subtracting the convolved image from the original LAB image. Square then add them all together and take the square root. That will give you an image which is the Delta E (color difference) between each pixel and it's neighbors. Would that be what you were thinking of?