PROJECT TITLE :

FastDeRain A Novel Video Rain Streak Removal Method Using Directional Gradient Priors

ABSTRACT:

The elimination of rain streaks from outdoor vision systems is an important problem that has lately been studied extensively. In this study, we provide FastDeRain, a new method for removing rain streaks from video that takes into account both the rain streaks' discriminative properties and the clean video itself in the gradient domain. On the one hand, rain streaks are scarce and smooth along the path of raindrops, whereas clean movies display piecewise smoothness along the rain-perpendicular path and continuity along the temporal path. The sparse distribution in the gradient domain is a product of these smoothness and continuity. To ensure anisotropic spatial smoothness, we use two l1 norms of unidirectional total variation regularizers and a 4 norm of the time-directional difference operator to reduce the density of the underlying rain streaks. Split enhanced Lagrangian shrinkage approach is used to solve provided model's minimization. It has been demonstrated that the proposed strategy is both successful and efficient by conducting experiments on both synthetic and actual data. Our solution outperforms existing state-of-the-art methods, especially in terms of running time, according to extensive quantitative performance measurements. Visit https://github.com/TaiXiangJiang/FastDeRain to get the source code for this project.