Deep Retinex image enhancement algorithm under weak Light Conditions

Abstract

An improved deep Retinex enhancement algorithm is proposed to solve the problems of large color deviation of reflection component and low detail of illumination component in low light condition. The Convolutional Block Attention Module (CBAM) is embedded in the enhanced network to extract the spatial and channel information of the image to improve the color distortion. Bilinear interpolation method was used to highlight details with the weight of adjacent spatial information. Finally, the enhanced image was obtained by merging R and L components pixel by pixel. The experimental results show that the subjective visual effect of the algorithm is more natural, and the objective evaluation indexes are greatly improved.