A SPARSE BASED RAIN REMOVAL ALGORITHM FOR IMAGE SEQUENCES

Ramya Chinniah and S. Subha Rani

Keywords

Sparse, dictionary learning, TOMP, rain removal

Abstract

Videos taken for border security and outdoor surveillance in bad weather conditions will severely suffer from degradation of information. This paper presents a rain degraded image enhancement algorithm based on sparse coding for robot vision. Sparse coding is a technique of finding a sparse representation for a given signal with a minimum number of significant coefficients corresponding to the atoms in a dictionary. Hence the size of the dictionary is usually a trade-off between approximation speed and accuracy. This paper adapts K means clustering based dictionary pruning algorithm to find an optimized dictionary for the given data set. This optimized dictionary selection will provide an increased convergence speed and performance to the proposed method by ensuring minimum error as well as sparsity of representation. The proposed method makes use of online dictionary learning and tree search based orthogonal matching pursuit (TBOMP) for sparse coding to retrieve the rain degraded image. This paper also examines the proposed method with other well-known dictionary learning and sparse coding techniques. The experimental results show that the proposed method provides improved performance in visual quality and also provides less computation time.

Important Links:



Go Back