Algorithms of Clustering and Classifying Batik Images Based on Color, Contrast and Motif

Veronica S. Moertini, Benhard Sitohang

Department of Informatics Engineering – Bandung Institute of Technology

Jl. Ganesha 10 Bandung 40132, Indonesia

Email: moertini@home.unpar.ac.id & benhard@informatika.org

An interactive system could be provided for batik customers with the aim of helping them in selecting the right batiks. The system should manage a collection of batik images along with other information such as fashion color type, the contrast degree, and motif. This research aims to find methods of clustering and classifying batik images based on fashion color, contrast and motif. A color clustering algorithm using HSV color system is proposed. Two algorithms for contrast clustering, both utilize wavelet, are proposed. Six algorithms for clustering and classifying batik images based on group of motifs, employing shape- and texture-based techniques, are explored and proposed. Wavelet is used in image pre-processing, Canny detector is used to detect image edges. Experiments are conducted to evaluate the performance of the algorithms. The result of experiments shows that fashion color and contrast clustering algorithms perform quite well. Three of motif based clustering and classification algorithms perform fairly well, further work is needed to increase the accuracy and to refine the classification into detailed motif. 

Download Article