multi-scale retinex is a luminance perceptual algorithm. it is widely used in image enhancement. to get best effect, the parameters need be adjusted according to the scene manually. in order to handle this issue, a novel retinex algorithm based on real-coded genetic algorithm and human visual system was proposed in this paper. the result was quantified and selected by a novel human visual system based comprehensive fitness function. the employed fitness function operates adaptively and tends to provide outstanding comprehensive results in both image clarity and color quality. images captured in low-light conditions were processed by proposed method. the result indicates that the new algorithm can improve image brightness and avoid color distortion at the same time. © 2016 ieee.