The authors look at texture analysis as an activity which simultaneously considers (i) a texture measure, (ii) a texture algorithm and finally (iii) misclassification properties. Because the structural characteristics of textural fields are difficult to define, an immediate consequence is the numerous number of texture measures required. Several segmentation algorithms are also required. The authors report some experiments in which examples from (i) and (ii) are simultaneously studied on a "Malaysian" data-set. In particular they use the entropy (E) anti contrast (C) from the co-occurrence matrix as texture measures, and simultaneously employ the idea of region growing and thresholding as a package in texture analysis.
展开▼
