Ancient Architecture Image Retrieval Based On Text And Content Multi-features

With the rapid development of digital imaging technology,digital image has become a significant data resource.It has become one of research hotspot that access to interest information from the massive amounts of digital images to domestic and foreign scholars as quickly as possible.The combination of digital image retrieval technology,which including text-based image retrieval and content-based image retrieval technology,and application it in ancient buildings image retrieval have important significance to the user access to ancient architecture Historic Background information quickly.The main work of this paper include:in the depth study of correlation theory and related technology about image color,shape and textural features applied block global color histogram and based on image retrieval technology of shape feature parameters and the texture primitive spatial distribution to extract images' color,shape and texture feature;based on ancient Chinese architecture image characteristics and respectively applied quadratic formula distance,Manhattan distance and the Euclidean distances to take similarity measure of color,shape and texture features by comparison each feature similarity measurement method;Through the traversal experiment of multi-feature fusion weight coefficient to 100 ancient buildings image obtained the optimum weights ratio of multi-feature fusion and establish a similarity measurement method of multi-feature fusion;One thousand ancient building pictures around the country were collected from the network and field,established image data set of small size ancient architecture and text data set of historical background information describes to a correspondence ancient architectures based on TEST image standard preprocessing.Finally,used MATLAB to design and realize image retrieval system of ancient buildings,and took image retrieval experiments to ancient buildings images,statistical analysis of the results of image retrieval recall and precision,drew PVR curve,the experimental results proves that the fusion multi-image retrieval method has better retrieval performance.This paper has a better grasp about image retrieval technology by reading a lot of related literature.On this basis,similarity measure of multi-feature weighting fusion were proposed by optimizing the texture feature extraction methods and applied to image retrieval.In MATLAB to achieve a systematic and retrieval experimental,test results show that this image retrieval method has a certain effectiveness and stability,and have made some achievements.