摘要
本文针对视频序列文字的检测与提取技术做了详尽的分析和讨论,内容包括:文字事件的检测、帧图像中文字区域的检测与定位。首先进行文字事件检测,然后进行边缘检测,加强了检测到的文字的边缘;对边缘进行尺寸限制过滤掉了不符合文字尺寸的边缘;进一步,提出像素密度δ的概念,并指出文字区域的像素密度δ应在某一阈值范围之内(δmin<δ<δmax)。通过像素密度δ滤去了非文字区域,应用投影法最终确定视频字幕所在区域。
In this paper ,comprehensive analysis and discussions are given on detection and extraction of texts from video sequence , in which several topics are involved : text event detection , text region detec- tion and orientation in video frames. An algorithm is provided which first detects text event and gets ed- ges, then makes a size restrict to the edges and wipes off non - text regions according to the textual ener- gy, after omnidirectionally expand the pixel of the edges lastly detects the remainder edges and detemins the textregions. The overlaying of detected horizontal edges and vertical edges enhances the text edges; the size restrict of edges helps to wipe off the non - text edges.
出处
《中国传媒大学学报(自然科学版)》
2008年第2期71-76,共6页
Journal of Communication University of China:Science and Technology