Application of Gray Level Co-Occurrence Matrix and Histogram Feature Extraction Methods for Batik Image Classification

Abstract

Batik is the art of integrating cultural elements like symbols and techniques into cotton and silk clothes. The Indonesian population has traditionally used batik in their daily lives. Each location has distinctive designs, patterns, and colors that reflect its meanings and philosophical perspectives. There are many different motifs on batik fabric, including geometric, geometrical, animal, and other designs. Batik motifs are frequently employed to convey social rank. The variety of batik designs and motifs is challenging for machine learning-based pattern detection and classification. This research applies the Gray Level Co-occurrence Matrix (GLCM) feature extraction method and Histogram feature extraction on batik images and the K-Nearest Neighbor (KNN) classifier. This study focuses on 4 batik patterns (motifs), namely Lereng, Nitik, Kawung, and Tambal. Dissimilarity, Correlation, Contrast, Homogeneity, and Energy from various angles and distances are the GLCM features employed, and their sum equals 1. Mean, standard deviation, smoothness, skewness, energy, and entropy are the histogram features employed. This work uses 120 batik image data—90 training data and 30 test data—. The findings indicate that at k=15 and k=17, accuracy attained using GLCM feature extraction is 77%, while Precision and Recall are 77%. Comparatively, the histogram feature extraction accuracy, Precision, and Recall are 53%, 54%, and 53%, respectively, with a value of k=27. This outcome demonstrates how feature extraction using GLCM can more accurately portray batik.