Research On Indexing And System Implementation In Content-Based Image Retrieval

Recently, with the rapid development of multimedia technology and computer networks, the capacity of digital images in the world is growing at a sharp rate. Whether military or civilian equipment, they will bring a daily capacity of several gigabytes of images, including images from satellite systems, various monitoring systems, scientific experiments, biomedicine and so on, and all of these digital images include a great deal of useful information. However, these images can not be accessed and used effectively in that they are disorderly distributing in the world. Therefore, how to retrieve required information from tremendous image data efficiently and rapidly has become a pressing issue.The traditional image retrieval techniques, which are based on text description and keyword, have many limitations, such as overload, inaccurateness and subjectivity and so on, and they can no longer satisfy the demand. In order to overcome the limitations of traditional image retrieval technology, we need to comprehensively, generally and objectively extract the content of images. So, content-based image retrieval (CBIR) technology emerges as the times require. The main idea of CBIR is to extract the feature vector of images according to the color, texture, shape and space object relationship information, and then build a warehouse of feature vectors. Afterwards, calculate the similarities between query images and images in the database, and then return the results. By using the objective physical characteristics of images, no or only a little manual intervention is required in CBIR. So, it is widely used in the occasions needing for automation.With the rapid increase on the capacity of images, the demand of real-time retrieval has become more and more important, and the response time for query has become the key issue for the practicality of image retrieval. And that the effective index schemes for high-dimensional data are key techniques for real-time retrieval in large-scale image database. Because of the speciality of image feature, traditional one-dimensional indexing schemes are no longer suitable for CBIR, we need to use spatial indexing techniques. This paper widely introduces the spatial indexing techniques, focusing on the study of R-Tree and VA-File, and then analyzes their characteristics and shortages. Subsequently, we combine the advantages of the both, proposing a new indexing method called RBVA-Tree, and the experiment shows the performance of RBVA-Tree is...