Abstract:

The traditional waveletdenoising is based on a selection of the threshold, but a new method based on statistical model in wavelet domain is presented, in which the prior statistical model is Gaussian scale mixture model. When the signal covariance is calculated, the image is cut into several blocks and the elliptic window is used, so the local covariance is achieved. Experiments results show that the performance of the proposed method is higher than the method which calculates the covariance in subband.