Texture Classification Based On Variants of Fundamental Units of LBP Using Complete Text on Indexes

摘要

This paper presents a novel approach for texture classification, generalizing the well-known local binary pattern (LBP) approach. The local binary pattern (LBP) descriptor is widely used in texture analysis because o f its computational simplicity and robustness to illumination changes. However, LBP has limitations to fully capture discriminating information. The uniform patterns derived on LBP have resulted with a medium classification rate. To overcome this cand to make best use o f ULBP, this paper proposed extended uniform patterns. In this paper, the variants offundamental units o f texture derived from uniform Local binary patterns are integrated with (extons and statistical features are derived on them fo r a precise texture classification. This paper initially transformed the raw texture image into extended uniform patterns (EUP), the complete texton indexes are derived on EUP and GLCM features are derived on EUP-CTM for efficient classification. This paper derived three types o f EIJP and on each o f this CTM is derived. The EU-CTM is a frameyrork, which consists in encoding both contrast information and texton patterns o f a 2 x 2 grid in a precised manner. Then, spatial relationships among the neighboring pixels are measures using GLCM. This allows producing a more discriminative encoding than several state-of-the art methods based only on intensity information. The proposed framework is compared with various methods and experimental results indicate the superiority o f the proposed schemes.