摘要
提出一种基于相位编组提取数字视频文字区域的算法.本算法首先基于相位编组抽取视频帧中的直线,然后过滤掉弱边缘直线;对边缘进行尺寸限制过滤掉了不符合文字尺寸的边缘.进一步提出了像素密度α的概念,并指出文字区域的像素密度α应在某一阈值范围之内(αmin≤α≤αmax),通过像素密度α滤去了非文字区域,最后应用投影法最终确定文字区域.以上方法的结合保证了本算法的正确率和鲁棒性.最后选用不同类型的视频素材对本算法进行了实验,表明本算法具有较好的正确率和较快的计算速度.
In order to extract text region in digital video, a text extraction algorithm is presented based on line extraction. Firstly, we extract lines in video frames based on organizing phases into groups and delete thin edge lines; the size restrict of edges wipes off the non-text edges; further more, we present the conception of pixel density a, and indicate that a is in a threshold scope（αmin≤α≤αmax）, according to a wipes off the non-text regions, eventually get text regions using image projection. The combination of these methods guarantees the performance of this algorithm. Our experiments show that the results of text extraction in digital video with this algorithm indicate that this algorithm has satisfactory performance of correctness and computing speed.
出处
《武汉大学学报（工学版）》
CAS
CSCD
北大核心
2005年第4期104-108,共5页
Engineering Journal of Wuhan University
关键词
数字视频
文字提取
相位编组
直线抽取
digital video
text extraction
organization of phases
line extraction
