Image denoising using a block-median pyramid
Vladimir P. Melnik, Ilya Shmulevich, Karen Egiazarian, Jaakko T. Astola · 2003
A block-median pyramidal transform, based on the median operation over non-overlapping blocks and linear Lagrange interpolation, is considered for image denoising. In addition to the soft and hard thresholding schemes, other techniques are employed. Firstly, the partial cycle-spinning algorithm is used to achieve near translation invariance. Secondly, local adaptation in the transform domain is used for adjustment of parameters in accordance with spatial behavior of the image.