Regularization Parameter Selection for TV Image Denoising Using Spatially Adaptive Local Spectral Response

Abstract

Currently, automatically estimating regularization parameter is still an open problem. To address this problem, an adaptive regularization parameter selection scheme is exploited for the total variation regularization (TV), by means of the local spectral response, which employs content-aware patches to calculate the spectral response of the TV procedure to locally estimate the regularization parameters in a content-aware way. Experiment results on simulated and real noisy image show the relatively good performance of our proposed selection scheme based TV, in visual improvement and peak signal to noise ratio value.