| With the rapid development of Internet and the explosive increase of multimedia resources in WWW, it becomes more and more difficult to look for interesting resource. In order to retrieve web multimedia effectively, this thesis researches the method of web-oriented multimedia semantic information extracting. The thesis summarizes the present methods of extracting semantic information of web multimedia, separates them into two categories: the method based on image content and the method based on web exterior information resource, and points out their merits and faults. The method of extracting semantic information of an image based on web exterior information resources is researched and realized. The semantic information of a web image is closely interrelated with its context. The interrelated texts of an image include: the filename, the surrounding text, the alternative text, the caption of the web, the caption of the linked web or the filename of the linked image, the image URL, the web URL, the linked web URL and the subject of the web. The semantic information of an image is contained in these interrelated texts. It is feasible to extract semantic information of an image from the texts. In this thesis, the method of extracting interrelated texts of an image and the method of transforming code, translating, segmenting words are presented. In order to extract semantic information of an image, semantic dictionaries such as topic dictionary, main body dictionary and attribute dictionary are built. And the arithmetic of extracting them is brought forward. Besides, the arithmetic of automatic appending topic words to the dictionary and the arithmetic of classifying image topic are presented. Based on the work, the web-oriented system of extracting semantic information of an image is designed. The system is divided into four parts: the part of meta search engine and control, the part of extracting texts and pretreatment, the part of automatic...
|
PDF Full Text Request