A Novel Wavelet Threshold Optimization Via PSO for Image Denoising

Abstract

Threshold selection is extremely important in wavelet transform for image denoising. The threshold selection problem can be viewed as continuous optimization problem. Recently, Particle Swarm Optimization was introduced to solve this problem, but its effectiveness is destroyed by the premature convergence. In order to overcome this drawback and obtain satisfactory effect, this paper proposes a modified chaos Particle Swarm Optimization algorithm for threshold selection, then adopts the optimal threshold achieved and a non-negative garrote function to process wavelet decomposed coefficients. When the premature convergence occurs, chaos search strategy will come into effect to help particles jump out of local optimization, and seek global optimization. Experimental results reveal the encouraging effectiveness of the proposed algorithm.