A Regularization Algorithm of Improved Data Fidelity Term and Weight Adaptive Bilateral Total Variation

Abstract

In view of the existing shortcomings of super-resolution that data fidelity term in MAP framework obtained from L1, L2 and other norm functions and super-resolution total variation (TV) or bilateral total variation (BTV) for regularization, Tukey norm function is used to construct fidelity term in the MAP framework which is combined with the weight adaptive BTV regularization term to improve the regularization algorithm based on TV or BTV, and a regularization super-resolution image restoration algorithm based on the improved fidelity term and the weight adaptive bilateral total variation (WABTV-Tukey) is proposed. Through comparative analysis of multiple simulation experiments, WABTV-Tukey norm and bilinear interpolation, BTV-L1 norm, BTV-L2 norm, BTV-Tukey norm, feature-driven prior MAP block, Lorentzian-Tikhonov Regular analysis, generalized total variation regularization based on neighborhood pixel expansion and other algorithms are used for comparative analysis. Subjective evaluation of visual analysis and objective quantitative evaluation of peak signal-to-noise ratio (PSNR) indicators verify the adaptability of WABTV-Tukey algorithm to different types of noise models, and the superior edge detail retention characteristics.