Storage And Indexing Technology Research And Implement On Image Search Engine

Along with the sudden growth of the World Wide Web as well as the rapid development of multimedia technology and the communication technology, retrieving and browsing the Internet multimedia information efficiently and effectively quickly becomes people's urgent demand. At present the existing image search engines, mostly uses the text key words based and the link information based technology to complete the image search and the image retrieval, but they has never used the visual content information of the image itself. So these systems'retrieval precision remains some certain limit. Meanwhile some content-based image retrieval systems that are only using information content of images for image retrieval are difficult to tackle the semantic gap.In response to these shortcomings, the study of Web images semantic content and visual imagery associated with the key feature of the integrated technology goes, and we are to develop a combination of visual image and semantic search engine named VAST to improve the accuracy of Web search images, and to provide information to users in diverse ways. VAST image search engine indexing system for the storage and visual information of the network images is an integration of semantic information indexing and retrieval of images and data management operation requested interface.Adaptive fuzzy mechanism is the based algorithm of the VAST image search engine storage and indexing system, and also is a key point to build the image index tree. Fuzzy clustering and dynamic multi-level indexing tree with dynamic, flexible, non-dependence on the characteristics of the target sample data, is a good mechanism to overcome the ambiguity of the cluster center C-mean sensitivity and the localized clustering of the network characteristics of the distribution of image data to solicit high ladder, the structure has better balance symmetry.The parallel storage architecture of the VAST image search engines indexing and storage system is another research focus. The parallel framework document adopted Linux physics investigating interfaces between systems and services to shield the actual physical storage system operation, when the operation of the system-issued document images available, the system image search system specific to the implementation of the core system resources. The index data structure that can be distributed storage, parallel processing, this would increase I/O speed, thus speeding up the retrieval speed.The system uses the JAVA realization network reptile to be partial and system through JNI technology transfer completed 16-dimensional image processing packages feature extraction and image frame integration. Parallel structure established by clustering storage system and C code, and some common code Shell achieve. System testing showed that vast visual images with frame accurate search and retrieval system to the more common images search engines rate increase, retrieval performance, good to meet user needs.