Low-light Image Enhancement Based on Joint Convolutional Sparse Representation and Adaptive Gradient Constraint

Abstract

To improve the visual quality of low-light images, we proposed a novel low-light image enhancement algorithm via joint convolution sparse representation and adaptive gradient constraint. The proposed method is based on the Retinex model. When estimating the illuminance image, under the joint representation of the directional convolution analysis sparse representation and the group local convolution composite sparse representation, the structure and texture of the illuminance image are constrained, and the optimized illuminance image is obtained. As for the reflection image, the adaptive gradient constraint is applied to suppress the noise and artifact and enhance the detail as well. Compared with the state-of-arts, the proposed algorithm has achieved great improvement in both subjective and objective evaluations.