Use of rank-conditioned morphological operators for texture classification

The texture analysis method developed has been successfully applied to a variety of natural textures. It is based on extraction of integer data, and the histogram feature data extraction is capable of implementation in hardware, based on a shift and compare algorithm on a pixel by pixel basis, comparison being undertaken over the chosen template shape size. Subject to the replication of the shape size area, this is highly amenable to parallel implementation down to the single pixel level. Comparison with the work of Wang and Haese-Coat (see Journal of Visual Communication and Image Representation, no.4, p.197-214, 1993) in terms of noise performance and discrimination indicate a less effective result. This is due in part to the non-optimal classification method adopted, and is compensated by lower computational intensity.