Color Correction on Multiple Scales

Rahman et al. (1999) extended the algorithm of Moore et al. They use a Gaussian to compute the blurred image and perform color correction on multiple scales. This method is not only used for color constancy but also for dynamic range compression. Output color o, is computed as 

This formulation is similar to Land s alternative version of his retinex theory where each term of the sum computes 

Finally, for display purposes, the output o must be transformed to the range [0, 1] or [0, 255]. Rahman et al. (1999) apply an identical offset and gain function to all pixels. In addition, Rahman et al. propose to multiply the output o, by the following factor 

---

Color correction on multiple scales, no gain factor. 

---