Image Retrieval Based On Combining Color And Texture Features Of Interest Region

With the rapid development of the Internet and digital equipments, the number of digital image database is growing at a shocking speed. How to effectively organize, manage and retrieve the large-scale image database becomes a very important subject in the field of image retrieval. Content-based image retrieval (CBIR)is the set of techniques to address the problem of retrieving relevant images from an image database based on automatically derived image features. In recent years, CBIR is a very active research direction. It has been applied in many fields.In this paper, lots of research work has been done around image retrieval technology about ROI-based Multi-features Comprehensive Retrieval algorithm.. Firstly this paper systematicly studys image feature extraction techniques, covering t color, texture, shape, spatial relations and semantic feature, and describes image similarity measure methods and retrieval performance evaluation.Secondly this paper detailedly studys and solves the multi-feature image retrieval algorithm, for example image multi-feature extraction algorithm, multi-feature Similarity Matching and multi-feature combination algorithm . In content-based image retrieval, global features are commonly used to describe the image content. On the one hand the problem with this approach is that these global features cannot capture all parts of the image having different characteristics and can't contain the spatial features. On the other hand, users could be interested in only object of images. In this context, global image descriptors are unusable. Therefore, local-features of image is necessary to improve retrieval result. Then this paper makes a special effort to study image retrieval algorithm based the region of interest, and adopt Harris algorithm to detect the points of interest to determine the region of interest. Based on interest points a novel algorithm using mutil-features to search images is presented. It utilizes the local color and texture features to express the contents of images. It not only overcomes the shortage that a single image feature can not be truly characterized, but also has a strong robustness to rotation, translation and criterion. Experiments show that the algorithm significantly improves image retrieval accuracy.Finally this paper detailedly study relevance feedback technologies based content-based image retrieval. On the basis of retrieval model, Relevance feedback algorithm is approximately divided into: Based on the query vector-point moving algorithm, Based on feature weight adjustment algorithms, Based on traditional statistical learning theory algorithms, Based on machine learning theory algorithm and Based on memory- model of relevance feedback algorithm. Experiments show that it improves image retrieval accuracy to add relevance feedback algorithm.