A Total Variation Based Image Restoration Algorithm With Noise Detection

The technology of image restoration is used in various fields and become the research object in the subjects of computer science,medicine,engineering,etc.It has important theoretical value and practical significance.Image quality will decrease inevitably due to the factors such as nonlinear characteristics of sensors,the optical system aberration,kinds of noise interference and relative motion between image equipment and objects and so on.Noise not only results in missing the image features,but also makes image blur.Total variation restoration algorithm and fast numerical algorithm are researched to process impulse noise and Gaussian-impulse mixed noise.The L₀TV-ROAD algorithm based on noise prior restraint processes different image degradation model of impulse noise.Firstly the ROAD method estimate the possible impulse noise;and then the total variation restoration model is constructed based on the data consistency constraints to available points;and then the PADMM optimization algorithm is used for solving the proposed total variation algorithm.Compared with the state-of-the-art methods,the L₀TV-ROAD algorithm has better performance on denoising.For image corrupted by mixed Gaussian-impulse noise,a novel restoration algorithm named TV-ROAD is proposed.The TV-ROAD method takes a two-stage iterative framework:identification of pixels corrupted by impulse noise and total variation based image restoration.The prior information of the impulse noise level is introduced into the ROAD method for the accurate detection of noise pixels,and the L₂TV based method can remove Gaussian noise effectively.The proposed L₂TV-ROAD algorithm can achieve superior restoration result and preserve image detail features.A large number of experiments prove that the convergence and effect of image restoration of the proposed total variation restoration algorithm based on noise detection combined ROAD method with total variation restoration algorithm are better than typical restoration algorithms.