Image classification by pattern and structure features clustering

Abstract

An approach for decomposition of visual images by clustering and pattern classification by structure features is considered. Multilevel hierarchical clusters such as rectangles, closed regions and integrated areas are proposed. Hierarchically constructed fragments are material to form pattern structure features. To reduce the clustering algorithm complexity the tolerance coefficient and quality criteria for merging process are proposed. The results of pattern classification by structure features for some image groups by hand and automatic regimes are presented in the article. Hierarchical trees are got for different number of structure coefficients as well as for absolute and relative merging functions.