Content based image retrieval using signature based similarity search

摘要: Two of the main components visual information are texture and color. In this paper, a content-based image retrieval system (CBIR), which computes color similarity among images, is presented. CBIR set techniques for retrieving semantically-relevant images from an database based on automatically-derived features. One tasks systems comparison, extracting feature signatures every its pixel values defining rules comparing images. These features become representation measuring with other in database. Images compared by calculating difference to descriptors. Previously methods used global extraction obtain For example, several like color, shape extracted each image. descriptors obtained globally means histograms features; coarseness, contrast, direction; about curvature, moments invariants, circularity, eccentricity. approaches not adequate support queries looking where specific objects having particular colors and/or present, shift/scale invariant queries, position dimension query may relevant. suppose one there two flowers different colors: red yellow, describe as average orange. This description certainly semantic meaning Therefore, weakness observable. Region-based attempt overcome previous method limitations representing collections regions that correspond such flowers, trees, skies, mountains.