Performance and Comparative Exploration of Reconstructed Quality for An Iterative SRR Algorithm Based on Robust Norm Functions Under Several Noise Surrounding

Abstract

The multi-frame SRR (Super Resolution Reconstruction) algorithm has become the significant theme in digital image research society in the last ten years because of its performance and its cost effectiveness hence many robust norm functions (both redescending and non-redescending influence functions) have been usually incorporated in the multi-frame SRR framework, which is combined a stochastic Bayesian approach and a regularization technique into the unify SRR framework. Consequently, this paper thoroughly presents experimental exploration of an iterative SRR algorithm based on several robust norm functions such as zero-redescending influence functions (Tukey’s Biweight, Andrew’s Sine and Hampel), nonzero-redescending influence functions (Lorentzian, Leclerc, Geman&McClure, Myriad and Meridian) and nonredescending influence functions (Huber). This paper utilizes two standard images of Lena and Susie (40th) for pilot studies and fraudulent noise patterns of AWGN, Poisson, Salt&Pepper, and Speckle of several magnitudes are used to contaminate these two standard images. The comparative experiment has been done by thoroughly changing all parameters such as step-size, regularization parameter, norm constant parameter in order to obtain the maximum PSNR (peak-signal-to-noise ratio).