An iterative algorithm for two-dimensional digital least metric problems with applications to digital image compression

A correspondence between the problem of two-dimensional digital least metric (DLM) fitting and data detection in serially concatenated systems in digital communication theory is described. Nearly optimal detection algorithms based on previous advances in iterative detection/decoding are applied to the DLM problem for two applications in digital image compression. The first application is least squares halftoning of digital images. The second is near-lossless (i.e., error constrained) minimum-entropy image compression. In both applications the use of the iterative algorithm yields significant improvements, measured in terms of residual metric, relative to previously suggested approaches to the DLM problem.