3) ABSTRAK Untuk membantu proses pendokumentasian citra Batik, dibutuhkan sistem klasifikasi yang cukup handal dalam mengklasifikasi dan mengidentifikasi citra Batik. Salah satu kehandalan sistem klasifikasi yang dibutuhkan adalah invariant terhadap rotasi. Kehandalan tersebut dibutuhkan agar sistem dapat diaplikasikan untuk mengenali citra dari berbagai macam sumber, seperti internet. Kehandalan sistem klasifikasi tidak lepas dari kehandalan metode ekstraksi cirinya. Salah satu metode ekstraksi ciri yang invariant terhadap rotasi adalah LBPROT. Namun, LBPROT memiliki kekurangan yaitu mengabaikan karakteristik lokal dari kekontrasan atau nilai varian. Di lain pihak, Completed Local Binary Pattern (CLBP) dan Completed Robust Local Binary Pattern (CRLBP) memiliki ciri yang dapat merepresentasikan nilai varian lokal tanpa mengabaikan struktur spasial lokal, yaitu ciri magnitude-nya, CLBP_M dan CRLBP_M. Oleh karena itu, pada penelitian kali ini diusulkan metode klasifikasi yang invariant terhadap rotasi, dengan menggunakan metode ekstraksi ciri yang menggabungkan kelebihan metode LBPROT dan CLBP_M (rotCLBP_M), atau LBPROT dan CRLBP_M (rotCRLBP_M). Hasil ekstraksi ciri akan menjadi data masukan untuk sistem klasifikasi Probabilistic Neural Network (PNN). Kinerja sistem diukur menggunakan akurasi. Hasil uji coba menunjukkan bahwa sistem klasifikasi dengan metode ekstraksi ciri rotCRLBP_M, lebih unggul dibandingkan dengan metode rotCLBP_M. Sistem klasifikasi dapat mencapai akurasi maksimal sebesar 90.34% untuk dataset Batik. Sedangkan pada dataset Brodatz, sistem klasifikasi dapat mencapai akurasi sebesar 87,92%.
This paper describes a novel method for extracting features of batik images. This method is called enhanced microstructure descriptor (EMSD). EMSD is the enhanced model of micro-structure descriptor (MSD) which proposed by Guang-Hai Liu. Different with MSD that uses only edge orientation similarity for creating micro-structure map and then utilises this map along with color values; EMSD adds a new micro-structure map that is based on color similarity and then utilises this map along with edge orientation values. The combination of MSD and the additional micro-structure descriptor is used as feature extractor in EMSD. This method is tested on 300 batik images, Corel datasets with 5,000 images and 10,000 images. We also compared EMSD to MSD and multi-textons histogram (MTH), which EMSD performance is superior than the other two.
Batik is a symbol reflecting Indonesian culture which has been acknowledged by UNESCO since 2009. Batik has various motifs or patterns. Because most regions in Indonesia have their own characteristic of batik motifs, people find difficulties to recognize the variety of Batik. This study attempts to develop a system that can help people to classify Batik motifs using Multi Texton Histogram (MTH) for feature extraction. Meanwhile, k-Nearest Neighbor (k-NN) and Support Vector Machine (SVM) algorithm were employed for classification. The performance of those classifications is then compared to seek the best classification method for Batik classification. The performance is tested 300 images divided into 50 classes. The results show the optimum accuracy achieved using k-NN with k=5 and MTH with 6 textons is 82%; however, SVM and MTH with 6 textons denote 76%. According to the result, MTH as feature extraction, k-NN or SVM as a classifier can be applied on Batik image classification.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.