My goal is to implement efficient image illuminance (or lightning) normalization algorithm. This algorithm is intended to process non-uniformly illuminated images (scans).
I think of trying two different approaches 1) Calculate I(x,y) = Mean(I)/(c*Std(I)); where I(x,y) - pixel value, Mean(I) - mean value of the pixels in image, Std(I) - standard deviation and c - some constant
2) Do gaussian blur of an image, then subtract blurred pixel from original one and multiply by some constant
Does anyone know if IPP has some more efficient algorithm?
Maybe your problem is solved using a combination Shadow/Highlight. Those functions are not simple histogram tweak functions, but rather complex frequency code to isolate pixel candidates who are dark or who are light, and then adjusting those, without adjusting others. Available in Photoshop CS3 and up.
I also would like to get to know more about a useable algorithm/implementation.