Low-Light Image Deraining Based on Higher Order Variational Model

Abstract

The existing rain streaks removal methods provide better deraining results, but it cannot be implemented on the low-light images due to the poor visual quality. To solve this problem, this paper presents a novel rain streaks removal approach using m fold infimal convolution of oscillating TGV(\(ICTGV^{osci}\)) regularization and Retinex theory for low-light images. Experiments on a number of challenging low-light rainy images are presented to demonstrate the efficiency and the flexibility of the proposed approaches in comparison with state-of-the-art methods.