An image search system with analytical functions for 3D color vector quantization and cluster-based shape and structure features

The application area of image retrieval systems has widely spread to the WWW and multimedia database environments. This paper presents an image search system with analytical functions for combining shape, structure, and color features. The system pre-processes an image segmentation from hybrid color systems of HSL and CIELAB. This segmentation process includes a new mechanism for clustering the elements of high-resolution images in order to improve precision and reduce computation time. The system extracts three features of an image which are color, shape and structure. We apply the 3D Color Vector Quantization for the color feature extraction. The shape properties of an image which include eccentricity, area, equivalent diameter, and convex area, are analyzed for extracting the shape feature. The image structure is identified by applying 2D Forward Mirror-Extended Curvelet Transform. Another distinctive idea introduced in this paper is a new distance metric which represents a semantic similarity. This paper has evaluations of the system using 1000 JPEG images from COREL image collections. The experimental results clarify the feasibility and effectiveness of the proposed system to improve accuracy for image retrieval.