Fast algorithm of image enhancement based on multi-scale retinex

Abstract

In this paper, a fast image enhancement algorithm based on multi-scale retinex in HSV colour model is presented. The proposed algorithm produces the result similar to the one which uses a nonlinear processing in the HSV colour model, but with less computational cost. It uses linear dependencies of RGB colours from the V-component of HSV model. Additionally, to speed up the images processing and enhance the local contrast is suggested to perform multi-scale retinex (MSR) computation only in the low-frequency area obtained by the wavelet transform. Experimental research was performed on more than 100 colour images having non-uniform brightness. Different algorithms based on retinex technology were implemented and their performance was compared. The proposed way of output image colour formation allows reducing processing time by 30%-75%, depending on the image size. The experimental data show that the usage of the wavelet transform in proposed MSR algorithm additionally leads to 2-2.8 times increase in processing speed.