改进的累积帧实现镜头转换检测技术

Improved algorithm for shot transition detection based on cumulative frame

镜头检测方法是视频内容分析和视频检索中的关键技术之一。对累积帧的检测镜头转换的方法提出了改进,包括:采用图像块颜色均值差作为前景后景判决参数,更好地避免了噪声影响;提出局部全搜索块匹配算法准确区分物体运动与渐变镜头。对镜头转换判决阈值的设定进行了修正。通过实验对算法检测效果进行分析比较,结果表明,改进后的算法能很好地解决累积帧方法对快速运动场景容易出现误判的问题,具有更好的镜头检测效果。

Shot detection method is the one of the key technologies for the analysis of the video content and video search.This paper detects the shot boundary of the video based on frames accumulation and proposes an improved algorithm,that＇s included： The algorithm makes use of the difference of the image block mean color value ,better to avoid the effect of noise.The algorithm proposes the local full search method for the block-matching to distinguish the motions from the dissolves.In addition,the paper also revises the threshold.The experience results show that this technique has a better efficient.