Wavelet based Multilevel Sub-Band Adaptive Thresholding for Image Denoising using Modified PSO Algorithm

The image de-noising is one in all the foremost studied areas within the field of image processing. There are many ways (like communication channel, imperfect sensors, interference etc.) by which the noise may affect the image. Depending upon the nature of noise and the image many techniques has been already proposed. However for any technique it is difficult to operate on different level of noises over different kind of images (like SAR images, Xray images, Ultrasound images etc.). The best possible solution for such cases is to use adaptive techniques. In this paper we are presenting a multilevel wavelet decomposition based adaptive thresholding technique which utilizes the modified Particle Swarm Optimization (PSO) algorithm to find out the optimal values for thresholds and level of decompositions for given objective function. The modification of PSO is done through random perturbation in particle velocities which induces small randomness in new particle position estimation. This randomness can effectively increase the particle search space, which ultimately provides a much better solution than the conventional PSO. Finally the proposed algorithm is validated by testing it over different kind of images corrupted by different values of noise.