| Summary: | 碩士 === 國立中正大學 === 資訊工程研究所 === 88 === Multimedia information systems are becoming increasingly important with the advent of broadband networks, high-powered PC’s and workstations, audio/visual compression standards, and many applications such as digital libraries, medical databases, trademark and copyright databases, geographic information systems, and video-on-demand systems. Because visual media data require a large amount of memory and computing power for storage and processing, it is greatly desired to efficiently index, store, and retrieve the visual information from image database systems. In this study, efficient indexing and searching strategies for feature-based image database systems using fast clustering and genetic algorithms are proposed.
Here each query or stored image is represented by a set of features (a feature vector) extracted from the image. A suitable similarity measure (distance function) is utilized to evaluate the ranks of the retrieved images. A set of fast clustering algorithms (a set of inequalities) is used to partition the Q images (feature vectors) in the image database system S = {x(1), x(2), …, x(Q)} into K clusters with K cluster centers {z1, z2, …, zK}. A query image (feature vector) first determines its C closest clusters among the K clusters (with respective to the K cluster centers). Then the most similar images for the query image are determined within the C closest clusters using genetic algorithms. Good simulation results show the feasibility of the proposed approach.
|
|---|
