Separability Based Tree Structured Local Basis Selection for Texture Classification

Kamran Etemad, Rama Chellappa

Abstract: A new algorithm for task dependent selection of wavelet packet trees for signal classification is suggested. The algorithm is based on a class separability measure rather than energy or entropy. At each level the class separabilities obtained from a parent node and its children are computed and compared. The decomposition of the node (or subband) is performed if it provides larger separability. The suggested algorithm is tested for texture classification. The method can also be used with other tree structured local basis e.g. local trigonometric basis functions. Also it can be applied to detection, classification or segmentation of different l-D and 2-D signals.<