基于运动目标特征的关键帧提取算法

针对运动类视频特征不易提取且其关键帧结果中易产生较多漏检帧的问题，提出基于运动目标特征的关键帧提取算法。该算法在强调运动目标特征的同时弱化背景特征，从而防止由于运动目标过小而背景占据视频画面主要内容所导致的漏检和冗余现象。根据视频帧熵值将颜色变化明显的帧作为部分关键帧，对颜色未发生突变的帧根据运动物体的尺度不变特征变换（SIFT）获得帧内运动目标的特征点；最后分别根据帧熵值及运动物体SIFT点分布提取视频关键帧。实验表明该算法所得关键帧结果集不仅漏检率较低且能够准确地表达原视频内容。

Key frame extraction algorithm based on feature of moving target

Motion features are difficult to extract which easily leads to missed and redundant frames in the result. In order to solve this problem, this paper proposed a method of key frames extraction based on feature of moving target. The method reduced redundant rate and missed rate of the result set by emphasizing the features of the moving target and weakening the background features of the frame. This method took a frame with a burst entropy change is taken as part of the key frame firstly. Then, it extracted the SIFT points of the moving target from the frame of which the entropy value had not suddenly changed. Finally, it extracted key frames according to the entropy and SIFT distribution respectively. Experimental results show that the miss rate of this algorithm is low. At the same time, key frame results can accurately and completely describe the main content of the video.