A simple illumination map estimation based on Retinex model for low-light image enhancement

Abstract

This paper proposes a effective illumination map estimation based on Retinex theory for low illuminance image enhancement. Firstly, initial illumination map is calculated by finding the largest element value in the b, g and r channels. Secondly, we adopt anisotropic filter operations to process initial illumination map. Then, we propose an adaptive gamma correction to process it to make the illuminance map more accurate. Finally, we adopt unsharp masking to enhance details to get our result. Objective and subjective evaluation illustrate the superiority of our proposed algorithm.