Transform coding of monochrome images with a statistical design of experiments approach to separate noise

A new transform coding technique based on a set of orthogonal polynomials has been proposed in this paper. In the transformed domain, statistical design of experiments approach is used to separate the spatial variation due to discriminable low level features (signals) from the spatial variation due to an unexplained source called noise. The proposed polynomial transform has a low computational complexity because it is configured as an integer transform. The proposed transform also has very good compression ability or separationability. The degree of this separationability can be controlled by specifying the level of the signal-to-noise ratio.