Pattern analysis and texture classification using finite state automata scheme

Abstract

The paper proposes a complete modelling of finite state automata along with the associated classifier for texture classification. Pattern analysis of the texture image is performed by proposing a symbolic pattern-based algorithm. This algorithm is developed based on the symbolic dynamics and finite state automata theory for estimating the state transition of the texture variations. Texture image is divided into several partitions, i.e., texture, background of the texture, shadow of the texture, etc. Finite automata state transitions are used to extract the features from the symbolised image. A binary classifier is designed to classify the texture categories based on the feature extraction from the finite automata theory. Pattern analysis is performed on the KITH-TIPS dataset for ten varied categories of texture. 99.12% classification accuracy is achieved when compared with other state-of-art techniques. The experimental study shows the better efficiency of the proposed system when compared to other existing methods.