Classification of Batik Motifs Using Multi-Texton Co-Occurrence Descriptor and Binarized Statistical Image Features

Abstract

This study aims to enhance the classification accuracy of batik motifs through a novel integration of Multi-Texton Co-Occurrence Descriptor (MTCD) and Binarized Statistical Image Features (BSIF). The primary objective is to develop a robust feature extraction method that effectively captures both textural and statistical properties of batik images, specifically utilizing the Batik Nitik 960 dataset. Our methodology employs a combination of MTCD and BSIF, followed by Principal Component Analysis (PCA) for dimensionality reduction, optimizing the model's ability to learn from diverse characteristics inherent in batik motifs to augment the diversity and robustness of the training data, we enhanced the Batik Nitik 960 dataset by applying vertical flipping, in addition to existing rotations. We explored three feature fusion approaches: Combination 1, where features are combined before normalization and PCA, achieving an accuracy of 99.948%; Combination 2, where normalization occurs prior to feature combination, also achieving an accuracy of 99.948%; and Combination 3, which applies PCA separately to each feature before combination, resulting in an accuracy of 99.896%. Experimental results demonstrate a remarkable accuracy in classifying these motifs, with the combined MTCD-BSIF features significantly surpassing the individual performances of MTCD at 95.729% and BSIF at 99.531%. This substantial improvement addresses the limitations identified in previous research, which reported an accuracy of only 0.71 on the same dataset. Furthermore, we explore the impact of various feature fusion techniques on classification performance, providing insights into the effectiveness of our proposed methods. Our findings suggest that the combined MTCD-BSIF approach can serve as a benchmark for future studies aiming to enhance classification accuracy in similar domains, thereby contributing to advancements in automated classification systems and their applications across various fields.