ADAPTIVE IMAGE FILTERING.

Abstract

The adaptive image filtering considered in this study includes a Kalman filter for noisy image enchancement and a generalized likelihood ratio technique to detect and estimate the jumps corresponding to object boundaries. The filter is adjusted when the jump is detected. When the transition matrix of the filter is unknown, it is determined by a method of simultaneous on-line estimation of parameters and states. Both the mathematical analysis and computer results are presented in detail. The procedures involved are highly effective and flexible, and computationally efficient.