Klasifikasi Citra Batik Menggunakan Co-Occurrence Matrices Berbasis Wavelet Filter

Abstract

Batik is the result of cultural arts that contains a philosophical meaning in each of its motifs. Various types of batik motifs create complexity in the recognition of batik image patterns. Classification of images into certain classes is also a problem in the field of pattern recognition. Machine learning is a method that is very developed at this time. Machine learning method is used to identify batik motifs through batik image classification. This study focuses on the image dataset of written batik which has two motifs, namely classical motifs and contemporary motifs. This study shows the experimental results of batik image classification using the Backpropagation Neural Network, Support Vector Machine and k-Nearest Neighbor classification methods. Co-occurrence matrices as wavelet filter-based feature extraction are used for input into batik image classification. The experimental results show that k-NN gets the best accuracy value of 95.56% while BPNN gets an accuracy value of 85.40% and SVM gets an accuracy value of 76.51%. Based on these results, it can be concluded that k-NN is the best method for classifying batik images with co-occurrence matrices as wavelet filter-based feature extraction.