Application Research Of Particle Swarm Optimization In Fuzzy Image Restoration

Blurred images are generally caused by out of focus of a camera, the relative motion between a camera and the original scene, and other factors like atmospheric turbulence effect in image acquisition process. This phenomenon is known as image degradation. Image blurring is a main handicap in visual information acquisition and subsequent image analysis and interpretation. Therefore, the issue of blurred image restoration has become one of the most critical and basic problems in the digital image processing. According to blurring mechanisms, blurred images are generally divided into three types including defocus-blurred images, motion-blurred images and atmospheric turbulence-blurred images or Gaussian blurred images. This dissertation only focuses on defocus-blurred images and motion-blurred images. Blurred image restoration is to solve the problem of image quality degradation and make the image as close as possible to the real scene. Generally, the point spread function(PSF) for a blurred image needs to be estimated firstly, and then an appropriate restoration algorithm is adopted to recover the latent image using the estimated PSF. Thus PSF estimation plays a paramount role in image deblurring processing, and it is the central focus of this dissertation. Wherein the main work is listed as follows:The related theories about image deblurring, such as image degradation model,the imaging mechanism and PSF of defocus blurred images and motion-blurred images are discussed firstly. The restoration algorithms including inverse filtering,Wiener filtering and Lucy-Richardson are given and compared through experiments.Secondly, the basic theories of particle swarm optimization(PSO) algorithm are discussed, which include the background of PSO algorithm, the basic PSO algorithm,the two models of PSO algorithm and the improved PSO algorithms.Then, a parameter estimation method based on grayscale mean gradient(GMG)and PSO algorithm is proposed to estimate the blurring parameter. The GMG valuesare taken as the fitness function values of the PSO algorithm, based on the property that the definition of an image is positively varied with its GMG value, so that a particle with maximum fitness is found, and thus the corresponding PSF is taken as the final result of estimation. Experimental results show that the proposed algorithm can accurately estimate the PSF parameters.Finally, Radon transform, the cepstrum analysis and some other classical PSF estimate methods are discussed and compared with the proposed parameter estimation method based on GMG and PSO algorithm and then adopted to restore a latent image from its blurred image, wherein good performance is shown.