Research And Application Of Automatic Image Annotation

As the development of the technology of network and multimedia, multimedia like image is developing very fast. People's demand of multimedia data such as graphics, image, video etc. is more and more intensive. However, traditional retrieval technology like image retrieval based on text can not satisfy the retrieval of multimedia information. Though the retrieval based on content solves the problem that people can retrieve by vision of image, it can not solve high level semantic image retrieval completely. Then it is the best way to establish semantic description and retrieval mechanism of image.How to extract the high-level semantic from visual characteristics, is the key question of Semantic-based image retrieval. According to sources of high-level semantics, there are three semantic extraction methods, which are the external information source-based semantic extraction, the interaction-based semantic extraction and knowledge-based semantic extraction. The automatic image annotation is to automatically obtain the semantic key words of images from visual characters and to support the semantic level search. Many machine learning methods are introduced to this area as they are good access to the corresponding relationship between image visual features and the text.Based on the overview of the current status of image retrieval, the thesis does a deep discussion on image segmentation, visual feature extraction and the recent researches on automatic image annotation. A novel windows-clustering algorithm based on patterns reduction strategy is proposed to segment images. The thesis also presents a score-based model of automatic image annotation. A semantic-based image retrieval system (SIRS) is developed as the practice of our theory. Finally, we conclude the whole thesis. The difficulties, hotspots and problem solved in the next step of research are pointed out.