Image Restoration Using Optimized Weiner Filtering Based on Modified Tikhonov Regularization

In this paper, the optimization technique to estimate the regularization term of the Weiner filter is proposed. Normally, the noise term must be known prior to the calculation. However, in our approach this prior knowledge of noise term is unnecessary. The modified Tikhonov regularization is applied to control the regularization term of the Weiner filter for the best-restored image. This parameter will be changed following the degraded image although the regularization parameter of modified Tikhonov regularization is a fixed value. In our experiments, several sizes of Gaussian blur and noise are used to produce the degraded image. The Peak Signal to Noise Ratio (PSNR) is considered to measure the restored image quality. The results show that the noise regularization parameter of the Weiner filter can be adapted, and comparison results are superior to the traditional Lucy- Richardson algorithm and the regularized filter except the degraded images which have small blurring. Moreover, this method has also tolerance from the non-exact blur kernel.