Classification of coal images by a multi-scale segmentation techniques

The development of an automated and effective scheme for classifying various major maceral groups within polished coal blocks is detailed. The purpose of segmentation is to partition the images into various types of macerals. A multiscale approach to segmentation is characterized in which the outcome of each process at a given resolution is used to control the other process at the next resolution. At each level, segmentation is carried out by maximizing the a posteriori probability, which is achieved by a relaxation algorithm similar to Beseg's. The segmentation approximation over a hierarchy of resolutions is conducted to speed up the estimation process and to include large scale properties of each pixel.