摘要： 针对含有运动目标的视频采用单一阈值提取关键帧时易出现漏检或冗余的情况，提出基于镜头内容变化率的关键帧提取算法。根据镜头相邻帧差和与镜头平均邻帧差的变化特点定义镜头内容变化率，由该变化率确定关键帧提取的高、低阈值，实现自动提取一到多帧关键帧。实验结果表明，该变化率对视频帧的内容变化有较好的描述能力，提取的关键帧具有很好的代表性。

关键词: 视频检索, 关键帧提取, 镜头内容变化率, 镜头平均邻帧差

Abstract: A key frame extraction algorithm based on Shot Content Change Ratio(SCCR) is proposed against miss or redundancy when using single threshold to extract key frames from video with moving objects. Shot Content Change Ratio is defined according to the variation of the sum of shot adjacent frame difference and the average shot adjacent frame difference. High and low thresholds are determined by the ratio, and a single or several key frames can be extracted from a shot. Experimental results show that this ratio can better describe video frames’ content change and key frames extracted are representative.

Key words: video retrieval, key frame extraction, Shot Content Change Ratio(SCCR), average shot adjacent frame difference

中图分类号: 

• TP312