Classification of Natural Textures using Rule Based Rank Matrix

Abstract

This paper derives a new frame work for the classification of natural textures based on gradient rank vectors derived on a 2 X 2 grid. This paper identified the ambiguity in deriving ranks when two or more positions of the grid possess the same value. To attend this ambiguity and without increasing the total number of rank vectors on d positions this paper derived a rule based rank vector frame work. This paper replaced the 2 X 2 grid with the column position of the Rule based Rank Word Matrix (RRWM). The range of column positions will be d! for d positions. This paper then divides RRW texture image, into a 3 X 3 grid and derives cross and diagonal rule based rank words. From this, the present paper derived Rule based Rank Word-Cross and Diagonal Texture Matrix (RRW-CDTM) and derives GLCM features for effective texture classification. The experimental results on various texture databases revels the classification accuracy of the proposed method. The proposed method is compared with the state of art local based approaches.