Rain Removal in Image Sequence Using Sparse Coding

Abstract

One of the major applications of image processing is robot vision. In this paper a rain degraded image enhancement algorithm is proposed, which is one of the applications of robot vision. The objective of the proposed method is to enhance the image sequences degraded by rain using sparse coding. Most of the other methods that deal with rain removal from image sequences are carried out only on continuous frames where temporal correlations among successive images are exploited. In sparse representation, with only a few dictionary elements, compared to the ambient signal dimension, can be used to well-approximate the signals. The proposed method makes use of Enhanced K-SVD (EK-SVD) for dictionary learning and orthogonal matching pursuit (OMP) for sparse coding to retrieve the rain degraded image. This dictionary selection will provide an increased convergence speed and performance to the proposed method by ensuring minimum error as well as sparsity of representation. In this paper the proposed method is also examined with other well known dictionary learning techniques. Simulation results show that the proposed method provides improved performance in visual quality and also provides less computation time.