Image Denoising Based on Wavelet Thresholding with Continuity and Self-Adaptability

This article proposes a new approach of thresholding aiming at achieving the best trade-offs between filtering the noises and retaining the image finenesses. A 3-order polynomial is utilized to interpolate on the hard-thresholds, enabling the continuity and differentiability for the new thresholding function. On the basis of Birge-Massart rules, the thresholds are quantized according to the actual condition of coefficients distribution at each decomposition layer, resulting with the thresholds possessing good local self-adaptability. The proposed thresholding is put to the test and the results suggest that the denoising effects are up to technical expectation and superior to that of other given conventional thresholding algorithms in terms of both PSNR and visual judgment.