利用时空特性的字幕检测与定位算法

Spatia-temporal Video Caption Detection and Localization Algorithm

提出一种视频字幕的检测与定位算法.利用视频字幕在时间上的冗余特性,以镜头为基本处理单元,采用监视-跟踪模型和扩展QSDD(PQSDD)度量来定位字幕的起始帧和终止帧,利用起始帧和终止帧确定起始字幕转换帧对和终止字幕转换帧对;对各帧对的差值图像利用边缘特性分别进行字幕定位,并提出一种基于背景复杂度的自适应阈值选取算法实现对边缘图像的二值化;最后对两幅差值图像定位出的字幕区域做逻辑与运算和连通区域分析得到最终的字幕区域.实验结果表明本文算法具有较高的检测速度和定位精度.

This report describes a spatio-temporal algorithm to detect and locate the overlay captions in digital video. The algorithm adopts the PQSDD measurement and a binary-search algorithm to decide the initiative and final frames including the same caption within one shot. Using the initiative and final frames, two flame transition pairs and two difference images are obtained. Caption region localization is subsequently applied by analyzing edge maps of the two difference images respectively. For getting binary edge images, an auto threshold selection algorithm according to region background complexity is proposed. Finally mathematic morphology and connective components analysis are performed to remove the false caption regions. Experimental results prove that the algorithm possesses higher detection speed and localization precision.