Study On Content-Based Of Multi-Feature Assembled Image Retrieval

Due to the steady growth of computer, multimedia, and Internet techniques, a huge amount of images are available. Currently, rapid and effective searching for desired images from large-scale image databases becomes an important and challenging research topic. 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 this dissertation, lots of exploratory research work has been done around some key techniques of CBIR, which include creating image database, technique of multi-feature assembled statically and non-linearly. The presented study is the current research focus of image processing and information retrieval. Thus, the research has both the theory and the application value.The main works of this dissertation are summarized as follows:Firstly, create the image database named XMLIDB. Object-oriented modeling approach is used to translate image into some meta-objects, and image database is modeled a hypermedia-net, in which nodes are images and links are the relation between images. The two-layers architecture is conductive to the expansion of the image feature and to increase the support of semantics.Secondly, build an image feature named Connection number. The method classifies the image into three set: domain color set (same set), border color set (indefinite set) and noise color set (contrary set). An image connection number model is proposed. Based on the model, the statistic of global shape and local color is used to form the connection number vector of image, with which we can query the image database.Thirdly, image retrieval method based-connection number is advanced. The image retrieval algorithm is developed basing on connection number vector. It is shown through performance experiments that this algorithm is of scale invariance and rotation invariance to some extent.