Rain Removal of Single Image Based on Directional Gradient Priors

Images taken on rainy days often lose a significant amount of detailed information owing to the coverage rain streaks, which interfere with recognition and detection intelligent vision systems. It is, therefore, extremely important recover clean rain-free images from images. In this paper, we propose removal method based directional gradient priors, aims retain structural original image greatest extent possible while removing streaks. First, solve problem residual basis sparse convolutional coding model, two regularization terms are proposed constrain direction stripe. Then, for layer in prior terms, multi-scale dictionary is designed detect stripes different widths. Finally, obtain more accurate solution, alternating multipliers (ADMM) used update coefficients alternately rainless rich details. experiments verify that algorithm achieves good results both subjectively objectively.