Blind image restoration based on automatic blur identify and total variation minimization

Abstract

In this paper, an adaptive blind image restoration algorithm is proposed. According to the feature that the certain blur may lead to the specific component distortion in the cepstral domain, we develop an automatic algorithm using Fourier-Mellin transform to classify and identify the point spread function (PSF) with cepstrum: for the usual types of blur, such as linear motion blur and defocus blur, we restore it with Wiener filter; while for others, we propose an improved total variation (TV) blind restoration algorithm. The algorithm combines the typical methods and blind methods of image restoration, achieving the adaptive blind image restoration. Experimental results show that it is effective in restoring degraded images under different environments, and it improves the restoring performance significantly under the presence of high noise level.