Performance Improvement of Average Based Spatial Filters through Multilevel Preprocessing using Wavelets

Image denoising filters intended to remove Gaussian noise, principally exploit a procedure called spatial averaging. Quite a lot of averaging approaches have been developed and numerous fall in the class of either pixel-based or patch-based or diffusion-based approach. While the designed filters get rid of the noise, the high frequency information will also be degraded, as the filters fit into a nature of integration. To preserve the high frequency information and hence the denoising performance, we propose a preprocessing filter designed in the wavelet domain, can be placed prior to the given existing spatial domain averaging filter. The proposed filter enhances high frequency information of given noisy image and obviously this enhanced information will also be degraded at some extent by the subsequent spatial domain filters. Accordingly, proposed preprocessing filter and existing average based spatial domain filter on a whole gives improved denoising performance. Simulation experiments have been conducted and it is proved that the proposed preprocessing filter certainly improves the denoising results of existing standard spatial domain filtering such as Anisotropic filtering, Bilateral filtering, Non local means filtering and recently proposed Probabilistic non local means filtering in terms of peak signal to noise ratio (PSNR) and structural similarity index (SSIM).