The Research Of Web Image Retrieval Based On Object

With the fast development of Network technology and multimedia technology, the digital images of the network are getting increasingly abundant. But it is always difficult for us to find what we need from the vast images. The image retrieval can resolve this problem. So, the retrieval of the Web image has become one of the hotspots of the multimedia information retrieval in recent years. The image retrieval technology includes the Text-Based Image Retrieval and the Content-Based Image Retrieval. It is the object in the image what people usually cares more in the Content-Based Image Retrieval. People can recognise the object in a picture easily, but it is almost impossible for you to recognise the main body information in all of the vast Web images manually, it is a tremendous task. In this paper, we build the objects'visual characteristic repository, based on the process of people's recognition of the object in an image. Through the extracting and the recognition of the visual characteristics of the objects in the images by computer, we recognise the main body of the image combining the repository, and then use it for the Object-based image retrieval. Our research includes the three main processes of the recognition such as the image analysis, the construction of the objects'visual characteristic repository, the object recognition. And we study the arithmetic of the Object-based image retrieval. The image analysis includes the image division, the extraction of the visual characteristic, and the correlative text extraction of the Web page. Here we use the edge division combining with the clustering division to divide images, and the division effect is better. We extract the color characteristic of the image object by histogram method, extract the texture characteristic of the image object by wavelet transformation, extract the form characteristic of the image object by invariant moments, use the BP Neural Network to recognise the texture and the form characteristic, and use the Chinese Natural language handling technology to extract the object name and the visual characteristic descriptor from the Web page where the image comes from. The construction of the visual characteristic repository of the image object is the base of object recognition. In this paper, we use the "and & or "method to construct...