A Robust Batik Image Classification using Multi Texton Co-Occurrence Descriptor and Support Vector Machine

Abstract

Batik is a unique pattern that symbolises characteristics and is a cultural heritage that is recognised by UNESCO. Batik is a research topic in image processing for image retrieval, object detection, pattern recognition, and classification. More and more new batik patterns and combinations between patterns are becoming increasingly difficult to recognise. Several studies have been proposed, such as KNN, Support Vector Machine, Convolutional Neural Network, SIFT, etc. to classify batik patterns. However, until now, it has not provided a reliable model proven from the accuracy that is still low. This study proposes the method of extracting Batik Image features using Multi Texton Co-Occurrence Descriptor (MTCD) with the Support Vector Machine (SVM) classifier validated with Logistic Regression (LR) to classify batik with high accuracy. The dataset used in testing uses Batik 300 and Batik 41k. The experimental results show that MTCD and SVM are a combination of very reliable techniques in classifying batik images. The accuracy obtained using SVM and LR is 1.0 and 1.0. Thus MTCD, SVM, and LR can be used to classify batik images effectively and reliably.