Simultaneous noise reduction and feature enhancement in images using diamond-shaped 2-D filter banks

A two-dimensional (2-D) multiresolution method for simultaneous noise reduction and feature enhancement in images is described. 2-D multiresolution analysis is carried out using a tree structure based on a two-channel, diamond-shaped 2-D filter bank for improved performance and processing efficiency. In this framework, a digital signal processing unit is embedded between the analysis filters and synthesis filters. The noise that contaminates the image is reduced by a technique similar to the wavelet shrinkage method recently proposed by David Donoho, while desirable features of the image are enhanced using a weighted synthesis of the subimages obtained from the multiresolution analysis.