Study On Key Techniques Of Region-Based Image Retrieval (RBIR) And System Design

Our world is dominated by visual information and a tremendous amount of such information is being added day-by-day. It would be impossible to cope with this explosion of visual data, unless the information is organized such that we can retrieve them efficiently and effectively. Content-based image retrieval (CBIR) is presented then for images. Existing CBIR systems are based on low-level visual features, so the bottleneck is the gap between these visual features and semantic concepts of images which is called semantic gap. Region-based image retrieval (RBIR) was recently proposed as an extension of content-based image retrieval (CBIR), which can narrow down the gap. In Region-Based Image Retrieval (RBIR), not all the regions are important for retrieving similar images. In retrieval, the user is often interested in performing a query on only one or a few regions rather than the whole image. The method is more efficient.In this paper, firstly, the developing process of RBIR is introduced, secondly, some technology for a region-based image retrieval framework are presented and then the results of retrieval experiments show it is a efficient system. Lastly, shortcomings and the direct of study in the future are putted forward.Image segmentation is the precondition of RBIR. An RBIR system automatically segments images into a variable number of regions using the Expectation-Maximization algorithm on combined features of each region, including color, texture and shape.Feature description is the key to the system. In this paper, color histogram, gabor wavelet and seven moment invariants as the features. We also improved the integrated region matching (IRM) to solve the matching of the regions.An efficient indexing of database and the method of decreasing dimensions optimize the system. In this paper, a set of lower and upper functions are applied to filter irrelevant images. Locally-Linear-embedding is used to decrease the dimensions of feature vectors. In addition, this paper also introduces the design of the system and the results of experiments.In order to evaluate this method, an image retrieval system and experiments are designed based on COREL database, for example, the lonely feature with integrated feature, the improved matching arithmetic with the SCAN, the improved indexing with the ordered indexing, CBIR with RBIR and so on. The experiments can be concluded that our method can retrieve at the level of objects and improve the retrieval performance.Although, now ,there it is not a successfully system of RBIR,a prototype has been developed to demonstrate the feasibility and efficiency of the system. In the end of this paper, I conclude the shortcoming of the system and put forward the dissertation with future research directions, for example, reference feedback, MPEG-7 criterion and so on.