Video Text Detection And Extraction Based On MPEG Frame

Text detection in video is based on text detection in image. Efficient indexing and retrieval of digital video data is an important aspect of video database management, such as key-word extraction, safe management of video content, video data miming, artificial intelligence, and so on. Because of it widely using, Text detection in video have becoming a hot Research focus in the image Extraction fields.Text detection in video is based on text detection in image, but has it unique nature. For example, Image resolution is not high, image luminance edges are faint. These will be operated by image processing method to reduce the bad impact. At the same time, compressed format, same text appear in multi-frame, relatively static text and motion background, they can be used to improve the detection efficiency, and it is our focus of the study. What is more, video indexing involves content analysis of video sequences, which is usually a computationally intensive process. Since most video data are stored in compressed format, processing directly in the compressed domain offers the possibility of computationally more efficient algorithms.Text detection method divide into two parts: text region positioning and text extraction. The first part is processing directly in the compressed DCT domain. Our algorithm makes use of the features encoded in compressed data to perform fast text detection. We initial position the text blocks using Edge detection algorithm, then the text blocks are assembled to text regions by morphological algorithm, we keep track of every region to record their standing time, delete the region if its standing time is too short. In the text extraction part, we extract the compressed data at first, this step will not reduce operation speed, and enhance the image by combination of multi-frame information which involves the same text region. Characters are reliably extracted by an adaptive threshold method after applying some noise reduction filtering in multiple frames. Such extracted characters can be directly fed into a conventional OCR system for recognition. Most existing methods for text detection in video usually operate on raw pixel data and only output the location of the detected text regions in a single frame. Our processing directly in the compressed domain offers the possibility of computationally more efficient algorithms. This is the innovation of our algorithms. What is more, our algorithms combination of multi-frame information to enhance the candidate image regions.Experimental results on several video sequences show that the proposed algorithm is able to detect and extract text in MPEG video sequences with various scene complexities.