Luminance Based MSR for Color Image Enhancement

In this paper we will focus on color problems which occur if we utilize MSR as an image processing technique, and propose a modified Luminance Based MSR (LB_MSR) algorithm. There are two novelties in our algorithm. Firstly, we only process the luminance channel which is obtained by principal component analysis (PCA). PCA provides orthogonality between channels, thus allow the color to remain stable despite the modification of luminance. Secondly, we just add the convolution results of different scales instead of multiplying the convolution results of each color value and the different scale Gaussian. In this way, the color noise in the shadow/dark areas can be suppressed.