Text Recognition Of Video Image Oriented To Caption Frames

With the development of multimedia technology, video has become the people's access to the channels of information. Faced with massive video databases, how quickly and efficiently retrieve video has become a hotspot. The text captions in the video may provide a crawl for the video retrieve.The research is carried out from two aspects as caption frame detection and text location of video image. First, we judge whether contains captions frames according to the region gradient value; then use the method of proved text location algorithm to locate the subtitles region on the basis of captions frames; In order to ensure text recognition work normally, we choose the image binarization algorithm based on entropy to suppress the noise on the basis of analyzing several interpolation methods; finally introduce the OCR (Optical Characters Recognition) to identify the text regions. Experiment shows that the presented method can avoid the false areas. It is suitable for following character recognition.