Perbandingan Metode Segmentasi SOM dan Fuzzy CMeans pada Content-Based Image Retrieval Berbasis Warna

Abstract

Abstract— The difficulty of determining the right keywords to get the right image is the disadvantages of image search methods based on metadata keywords. Current technology developments lead to content-based image retrieval (CBIR). One feature of image content used for image retrieval is the color feature. To improve CBIR performance, this study investigated the comparison between SOM and Fuzzy C-Means segmentation methods. This segmentation method separates the foreground and background of the query image to get better CBIR performance. This study used Wang Dataset image database. Testing is done with test images that have undergone scale, rotation and blurring. The results of the test show that the segmentation method improves the recall value or the true image that has been found, but significantly reduces the precision value or the true image ratios of the overall images found compared without using segmentation method.