视频相似性分析指根据内容度量两个视频之间的相似度，在视频搜索、视频管理、视频分析、视频监控等有着重要的应用价值。度量视频之间的相似度，需要先提取合适的特征，但是现有特征提取方法多是在像素域进行，特征的计算和存储都需要较多资源，即便针对压缩视频的研究也多是针对MPEG标准进行。然而日益普及的H.264标准有着不同于MPEG1,2的新技术，因此已有压缩域方法不能直接应用于新编码标准。本研究旨在针对H.264/AVC压缩视频，挖掘压缩码流中的有用信息，从基于关键帧集和基于流间特征变化两个角度研究视频的特征表示及特征空间上的视频间相似性的度量方法。

Video similarity analysis is to measure the similarity between two videos based on video content. It plays an important role in the research area such as video search, video management, video analysis and video surveillance etc. Video similarity measurement needs to extract the proper features. But the majority existing methods extract the features in pixel domain, and the computing and storage will need more resources. Even for the research works in compressed domain, the methods are usually for the MPEG compressed video. However, the H.264/AVC standard is broadly adopted in most of the videos storage and transmission applications. Due to many new technologies are introduced by H.264 coding standard, the existing methods in compressed domain can not be directly applied to the new coding standard. This study is focused on the similarity measurement of two video clips based on the H.264/AVC compressed domain. On the basis of mining the useful information from the compressed domain, the effective feature representations for video contents are established and similarity measurement algorithms are designed. The study is carried on from two aspects, one is based on the set of key frames and the other is based on the feature change in the data flow.