Image filtering algorithm based on digital TV filter coupling with residual error correction

摘要

In this paper, we propose a digital TV (total variation) restoration model based on graph theory, motivated by the classical continuous TV restoration model of Rudin,Osher and Fatemi. We investigate its advantages in de-noising and enhancing non-texture images, and also analyze its shortcomings in deal with the texture images. Following the ideas of Yves Meyer in a total variation minimization framework of L. Rudin, S.Os-her and E.Fatemi, we propose a new image filtering algorithm subsuming the digital TV filter and the residual error correction scheme. In our algorithm, we decompose a given image /sub u//sup 0/ into a sum of two parts u + v by two-step method, where in the first step digital TV filter is used to obtain u which is the sketchy approximation of /sup u//sub 0/ , while the texture information in v is extracted by the residual error correction scheme using wavelet and wavelet packet thresholding technique. Experiments show our hybrid algorithm has more abroad application perspective and has better performance than digital TV filtering algorithm on real image.