Three dimensional biological microscopic image restoration with adaptive local regularization parameter based on wavelet domain

A method is proposed for adaptively choosing local regularization parameter based on gray scale difference estimation, and used to three-dimensional (3D) biological microscopic image restoration MPMAP algorithm. Every optical-sectioning image of 3D microscopic image with noise is decomposed in wavelet domain, and then its high frequency images of horizontal, vertical and diagonal direction are reconstructed. Then the images are convoluted respectively with corresponding direction operator, and the local gray scale differences of high frequency images before and after convolution are calculated. The minimum of corresponding local gray scale differences in the high frequency images is selected as local gray scale difference estimations of the optical-sectioning image, then the local regularization parameter of the optical-sectioning image is chosen by mapping local gray scale difference estimation. The local regularization parameter of the 3D microscopic image is made of the local regularization parameters of every optical-sectioning image. The test results show that the local regularization parameter based on gray scale difference estimation can describe more accurately intensity and position of noises than noise variance estimation. The local regularization parameter is used to 3D biological microscopic image regularization restoration with MPMAP algorithm. Experimental results show that better super-resolution effect is reached than whole regularization parameter MPMAP.