Image Noise Removal via Hierarchical Subbands Shrinkage Modified by Particle Swarm Optimization

An image noise removal method based on hierarchal subbands shrinkage (HSS) via smooth-shrinking method is proposed in this paper. Small coefficients are often thought to be due to noise, but some of them may be due to fine image features. Unlike the soft- or hard-thresholding just set those small coefficients to zero,smooth-shrinking shrinks them to some extent. Universal thresholding shrinks all subbands by a universal threshold, but HSS shrinks each subband by special selected thresholds. As an efficient optimization method, Particle Swarm Optimization (PSO) is adopted to modify or optimize those special thresholds for smooth shrinkage. HSS was superior to universal shrinkage and Wiener filering in terms of MSE. HSS could keep more fine details preserved and obtain better subjective image quality. The optimization achieved by PSO improved the noise removal performance of HSS.