Novel Framework for Texture Classification using SIFT Local Features
Đỗ Văn Tuân, Jongsoo Lee · ITC-CSCC :International Technical Conference on Circuits Systems, Computers and Communications · 2007
We propose an approach using the local features of the texture image for texture search and retrieval. The local features (referred to descriptors) are extracted from the texture image using Scale Invariant Feature Transform (hereafter, referred to SIFT) algorithm. In general, a texture can be characterized through textons, which are formed by clustering the local features. To generate the texton dictionary from the database, the adaptive mean shift clustering algorithm is executed on a randomly selected subset extracted descriptors from the database. Once the texton dictionary is built, the texton histograms can be created for each texture image. A framework for measuring the similarity among image textures using texton histograms is introduced. The proposed method is tested with a large database, and the quality results are given.