Research And Implement Of Index Storing And Searching Method In Image Retrieval

The main techniques about image retrieval include: feature extracting, feature Indexing and retrieval.The image retrieval based on the color histogram can easily get the Image index, but the color histogram only showes the statistic layout of the various colors, without the information of the space information of the image. Images with the different color layout can have the similar color histogram. So it will be possible to get widely different images with the similar color histogram, which is a big problem. The image retrieval based on the combination of the color and the space information can overcome the weakness. But in the general image retrieval method, it is hard to retrieval the index. MPEG-7 provides a standard named color layout, it shows the space information of the color and it can be used to get the indexing rapidly.In a CBVR System, the image is usually expressed with the form of a vector. So it will be a high pay to calculate the distance between the vectors one by one. Using the structure of tree in the clustering algorithm will be a good method. And two highly effective image retrieval algorithms based on CM-Tree were implemented. It uses the distance table saved in the node to estimate and cut the searching path, this advance the performance of the algorithm highly. The system uses CM-Tree as the structure for indexing. And there is no need to build indexing tree every time after storing the indexing tree in files. Storing the indexing tree is not so easy as the lining-structure, using the classical relationship of the relative database for storing seems to be a good solution. And this can avoid building the CM-Tree again.Finally, the techniques of index storing and feature retrieval are implemented in an image retrieval system. The conclusion can be gained that the techniques above provide the system with the good performance.