Optimal approach for neutron images restoration using particle swarm optimization algorithm with regularization

In this study, we want to implement a new approach to the nonlinear degraded images restoration problem which is useful for neutron radiography gray images enhancement. We attempt to reconstruct or recover an image that has been degraded, using some a priori knowledge of the degradation phenomenon. Our approach is based on using swarm intelligence optimization algorithms such as particles swarm (PSO) to solve a least squares minimization ill-posed problem. Many works have been done using intelligence techniques, ranging from neural networks, fuzzy logic and evolutionary algorithms, to swarm intelligence that we will try to use to minimize the Total Variation (TV). Instead of the standard Tikhonov regularization method which is most often used and to get smoothed images in presence of noise, a Laplacian constraint is introduced for regularization purposes. Using some image quality metrics such as Mean Square Error (MSE) and Peak Signal to Noise Ratio (PSNR), we can judge that PSO algorithm yields excellent results and good efficiency in noisy images restoration. These results indicate that the use of particle swarm optimization algorithm enables us to perform neutron radiography images restoration properly. © 2010 Asian Network for Scientific Information.