【摘要】 提出一种基于彩色边缘检测和游程平滑的视频文本提取方法 ,首先用彩色边缘检测算子检测出图像中的边缘 ,然后用多次水平和垂直的游程平滑操作使得文字边缘形成连通区域 ,并去掉部分孤立的噪声 ,再通过对连通域的分析和分解定位出文本区域的边界 ,最后对文本区进行确认 .实验表明 ,本文的视频文本自动提取方法具有较高的文本提取率和较准确的边界定位 .更多还原

【Abstract】 This paper presents a text extraction method based on color edge extraction and run-length-smearing. First color edge extraction operator is used to get the edge image, and then multiple run-length-smearing processes are used to get big, connective regions and remove some isolated noises. Next, after analysis and further segmentation on the connective regions, some candidates are obtained. At last, the candidates which passed a verification process become the final text regions. Experiments show that the method has high text extraction ratio and can get relatively exact text bound.更多还原