Hybrid Image Filtering Method Based on Wavelet Transform

The image acquisition process often ends with obtaining an image tainted with parasitic information which alters its quality. In order to analyze the image effectively and thus make reliable decisions, it is essential to reduce or even eliminate the noise that corrupts it. This work proposes a method of filtering images in order to reduce the noise they contain. The proposed method is a combination of bilateral filter, bivariate shrinkage and Bayes Shrink thresholding in the wavelet domain. This method consists of applying the bilateral filter to the noisy image and then decomposing it into eight wavelet sub-bands. Subsequently, bivariate shrinkage is applied to the first-level detail sub-bands, and BayesShrink thresholding is applied to the second-level detail sub-bands. The filtered image is obtained after wavelet reconstruction. The experiments were carried out on two most used images in the literature (Cameraman and Lena) corrupted with Gaussian with noise variance σ2 N = {0.05, 10, 20, 30, 50}. The performance of the proposed method was evaluated in terms of PSNR. The results obtained compared to certain approaches in the literature show that the proposed method is efficient and considerably outperforms those of existing works with an optimal PSNR value.