An algorithm for texture analysis is presented. The algorithm consists of a bank of filters, single Gabor (symmetric) filters and the combined Gabor (asymmetric) filters. The former are responsible for detection of the texture frequency, the micropattern size, and the orientation. The latter are used to segment the textures differing only in their micropattern structures. These two kinds of filters form a complex set of filters for texture segmentation. The combined filter is formed by three Gabor filters. One is called the main filter and the other two are subfilters. The two subfilters are Gabor filters shifted. The linear combination of the three filters forms a combined filter, an asymmetric filter. The wavelet transform of this combined filter forms a set filters which are able to "perceive" the structures of micropatterns in textures.
展开▼