Texture Representation Using Galois Field for Rotation Invariant Classification

Abstract

A novel method for rotation invariant texture representation using Galois Field is proposed in this paper. Rotation invariance is accomplished due to the commutative and associative properties of Galois Field addition. The bin values of the normalized cumulative histogram for Galois Field operated image are considered as texture features which are inherently rotation invariant. These features are used for texture classification; K-Nearest Neighbour classifier is used for classification. The Brodatz, Mondial Marmi, Outex and Vectorial datasets are considered for experimentation of the proposed method. The experimental results are compared with Rotation Invariant Local Binary Pattern (RI LBP) and Log-Polar transform method. It is observed that the proposed texture representation is more effective as compared to other two methods.