基于内容的图像搜索重排序算法研究-电子与通信工程专业论文.docxVIP

II II ABSTRACT With the rapid development of multimedia technology, a large number of digital images swarm into peoples daily lives. The overwhelming of multimedia data makes an urgent requirement to accurately and rapidly find the concerned images and videos. Traditional image retrieval based on text may cause “semantic gap”. Thus, image retrieval based on content is more and more important and image search reranking technology becomes the hot issue in this area. However, the high-dimension of the visual features usually causes the problem of “curse of dimensionality” and brings the high computational and storage burden. In this paper, we conduct a deep research on network image search reranking, and obtain following achievements: Firstly, we extracted features such as color, edge and histogram feature from initial retrieval results based on text. Secondly, we improve Locality Preserving Projections (LPP) by incorporating Pearson correlation coefficient and relevance degree information into it. It can enhance the accuracy of dimension reduction effect for image features. Finally, we utilize Ranking SVM to train a ranking model and rerank initial results. Dimensionality reduction process could not only find the internal structure of the data accurately, but also save storage space. Based on the above research, we study further about relevance degree information. We propose a simple reranking method based on pseudo relevance degrees which utilize the information of relevance degree. With few relevance degree labels，we calculate the pseudo relevance degrees of other unlabeled images, and then rerank images based on the pseudo relevance degrees. The proposed image search reranking method is model-free and does not require iterative so that it has low computational cost and performs effectively. KEY WORDS：Image Retrieval, Dimensionality Reduction, Relevance Degree, Reranking PAGE PAGE IV 目 录 HYPERLINK \l _bookmark0 目 录III HYPERLINK \l _bookmark1 第一章 绪论1 HYPERLINK