Several improvements for a wavelet denoising method
Wenli Xu · Journal of Tsinghua University(Science and Technology) · 2002
The recently proposed wavelet denoising method, locally adaptive window -based denoising using maximum likelihood (LAWML), can be further improved using three strategies proposed here. First, a threshold is used to eliminate small coefficients, which leads to smoother reconstruction and higher compression. Secondly, the size of the neighbor window was varied based on the energy distribution in different levels and different classes of subbands to provide more accurate estimates of the variance. Finally, a rectangular neighbor window was used instead of the square window. Experiments showed that using these strategies, the modified LAWML provided better performance than the original.