Binary Tree of Support Vector Machine in Texture Classification Problem

J. Liu, L. Xu, and B. Fei (PRC)

Keywords

Binary tree, SVM, texture, wavelet

Abstract

Because of the low efficiency of traditional methods that utilize support vector machine in multi-class classification problems, we present a new classification and training method named Binary Tree of SVM, decreasing the training and classification complexity to the great extend, while the complexity of the original problem is not increased. Compared with the old methods, Binary Tree of SVM is more suitable for classification tasks with many classes. In addition, based on the wavelet decomposition, this paper presents a texture features extracting method that uses the information within and between images of different decomposition levels. The experiments with the methods in this paper have achieved satisfying results.

Important Links:



Go Back


IASTED
Rotating Call For Paper Image