Experiments on the use of local statistics for adaptive image processing

In this paper, we report results from our experiments to find useful measures of local image statistics from small subblocks of data. Our criteria for the reliability of statistical estimates is that they should yield high quality results when used in adaptive image processing. We describe a method for estimating the correlation parameters of first-order Markov autocovariance models using relatively few computational steps. An application to block-interpolative data compression is given.