A Study of Texture Classification Using Spectral Features

摘要

Prior spectral feature study was not successful because of inaccuracy in 2-D spectral estimation. In this report, a generalization of Lim-Malik method is used to accurately estimate the ring and wedge spectral features for classification of textural images with a typical performance of over 80 percent correct classification for six texture classes. The results compare favorably with other texture features but require less computation using iterative algorithm. (Author)