A Content-Based Image Retrieval System Based On HTC Index

With the development of information technology and the progress of digital technology, everyone has cameras, smart phones or other equipment. More images produced by the way, image quality improved and the number of image is growth with astonishing speed. Faced with the rapid growth of the number of images, how fast, efficient and accurate image retrieval is the hot research focus now.The traditional image retrieval is text-based image retrieval. By the way of using a text to describe the image, the image retrieval will be converted to text retrieval. But describing the image by hand, not only wasting time, human resources or material resources, but also con not completely contain all the information of the image.Text-based image retrieval has so many problem, so there has been content-based image retrieval. In content-based image retrieval technology, image content description and index mechanism is the key research direction, it is also the difficult point in research. At the current computer vision and image understanding of the technical level, unable to realize the image semantic understanding. Image content mainly described by image low-level physical information, including color, shape,texture and so on.Content-based image retrieval system describe the image by its own information, like color, texture, shape or other information. Content-based image retrieval is objective. Index mechanism is the core of content-based image retrieval.Which include the establishment of the index structure, the query algorithm design and the result of the candidate, Faced with characteristics of high dimensional data,the query performance of the traditional index methods will below the sequential search, which is called ”dimension disaster”. In large-scale image database In order to improve the speed of image retrieval, it must adopt effective high-dimensional indexing mechanism.The main work of this article is that an image content retrieval system based on HCT(Hierarchical Cellular Tree) index structure is implemented, and the retrieval performance of system is tested and analyzed. The system is query by sample. The image description of the system use color, shape, and texture. The index of the System is HCT tree which is a measurement of dynamic access method and could insert data dynamically and has good balance. Aiming at the nucleus of original cells maybecome the nucleus of the new cells after cell division, this article put forward the improvement strategy and compare with the original method.According to the structure of HCT tree, the system use the approximate k nearest neighboring query method. This article Test and analysis the organization performance and retrieval performance of HCT tree by experiment. The result of the experiment show that build speed of the improvement strategy is better than the original method.