AN IMPROVED NAS-RIF ALGORITHM BASED ON THE WAVELET DENOISING AND IMAGE SEGMENTATION

摘要

An improved method is presented in this paper to overcome drawbacks of the original NAS-RIF algorithm. To begin with, the wavelet denoising technique is used to preserve the edge feature of the degraded image, restrain noise amplification, and increase the Signal-to-Noise Ratio (SNR); then, the image segmentation technique is applied in each iteration to find the precise support region of the object, where the non-uniform background is replaced by the mean of the background; the algorithm resetting of the convergence of the conjugate gradient is also employed here to speed up the convergence rate. The improved algorithm is experimentally shown to have better restoration effect and faster convergence rate.