Spatial adaptive wavelet thresholding for image denoising
Sheng Chang, Martin Vetterli · 2002
Wavelet thresholding with uniform threshold has shown some success in denoising. For images, we propose that this can be improved by adjusting thresholds spatially, based on the rationale that detailed regions such as edges and textures tolerate some noise but not blurring, whereas smooth regions tolerate blurring but not noise. The proposed algorithm is based on multiscale edge detection and image segmentation and then thresholding the coefficients of different regions with adaptive thresholds.