Improved Interscale-Based SURE-LET Approach to Image Denoising

Currently, the SURELET approach has been widely used for the removal of noise in corrupted image. However the original interscale-based SURELET approach rules out intrascale considerations of wavelet coefficients, which makes the denoising ability of the technique is reduced greatly. In this paper, the local Wienerlike estimator is incorporated into the original interscale-based SURELET thresholding function. The tests show the improved interscale-based SURELET gets the better denoising performance objectively and subjectively. Especially for Barbara image, the PSNR is on average improved by 0.52 dB compared to the traditional pointwise interscale-based SURELET approach.