【摘要】 视频数据中的文本可提供重要的语义信息。该方法首先用形态算子提取视频帧的高频分量,再通过区域标记、连通分量分析、形态后处理等一系列步骤对视频帧进行滤波,得到视频文本检测结果,最后再对其进行二值化处理,把文本字符从复杂背景中分离出来。这种算法能快速地定位文本区域,且不受文本颜色、语种、字体大小、排列方向、文本形变的影响。实验表明,该算法对视频文本进行检测定位精度较高,分割效果良好。更多还原

【Abstract】 Text in video can provide important semantic information. In this algorithm,firstly,the edge image is detected by morphological operators, in which the edge shape is fine kept. Then, the edge image is processed by area labeling, connected component analysis, morphological post-processing, etc. The final satisfactory results are acquired. This approach can detect various orientation, various changed shape text, which can detect not only caption text but also scene text. Experiments show that the algorithm can detect text region simply and effectively.更多还原