Image denoising using bivariate wavelet packet zerotrees and neighbor dependency
Pichid Kittisuwan, Widhyakorn Asdornwised · 2008
This paper presents image denoising methods performed within wavelet domain scheme by using wavelet packet zerotrees, and at the same time incorporating neighbor and inter-subband dependencies through NeighShrink and BiShrink [1] shrinkage functions, respectively. In particular, we call our proposed method as adaptive wavelet bivariate maximum a posteriori estimator (MAP) with NeighShrink threshold function namely, MAP_NBShrink_WP. In our second method, adaptive wavelet bivariate minimum mean square error estimator (MMSE) with NeighShrink threshold function, namely MMSE_NBShrink_WP, is proposed for image-denoising. Experimental results show that our proposed methods, MAP_NBShrink_WP and MMSE_NBShrink_WP, have better PSNR than BiShrink [1], NeighShrink [6] and BayeShrink [3] in oscillatory images (such as Barbara).