IMPLEMENTASI PRINCIPAL COMPONENT ANALYSIS (PCA) UNTUK TEMU KEMBALI CITRA MOTIF KAIN TENUN NTT BERDASARKAN WARNA DAN TEKSTUR
Abstract
ABSTRAK
Nusa Tenggara Timur is one of the island province that has a diversity of cultures in each region. One of them is a woven fabric with distinctive motifs that vary both in terms of color, shape and texture. To classify the motifs of woven fabric manually will experience difficulty and longer time. This research aims to to apply the CBIR system method for Retrieval of NTT woven fabric motifs. Content Based Image Retrieval (CBIR) is a image search method based on the similarity of color, shape and texture. Feature extraction methods used are color features using color statistics. Texture features using texture histograms. Principal Component Analysis (PCA) used in the feature classification process. To compare the similarity of the test image and the image in the database using the Euclidean Distance method. The number of sample images is 12, consisting of 36 images in the data base and 24 test images. Testing process is also done by adding noise salt & pepper and noise gaussian. Based on the test results, the level of success and accuracy of image retrieval up to 94,3% of the 12 samples tested. On the addition of noise salt & pepper the method still survives at the 0.05 density level of 61% and the gaussian noise at the mean & variant 0.01 level of 66.5%. In order to prevent a mis retrieval image outside the database, the application uses a thresholding technique.
Keywords: CBIR, Motif Kain NTT, PCA, Color & Texture