The Research And Application Of Video Text Extraction Approach Based On Gray And Edge

With the rapid development of communications technology, computer technology, broadband network technology and network technology, the number of digital video is growing at rapid pace. It have became a technical issue in urgent society that how to query the information in such big video database. Therefore, developing content-based video retrieval systems has a high practical value in the multi-media information field.The research of video retrieval system relates to text, audio, video processing, and text includes much semantic information, so the processing of text plays an important role in the research of video retrieval technology.Generally, the text extraction in the video includes three steps:(1)text detection and location, it determine whether the video frame contains text with the help of text feature, and if the video frame contains text, it is necessary to locate the place of text;(2)binaration, the purpose is to divide the text from complex background; (3)text recognition, nowadays, there are many mature third-party commercial software to complete its function, so the article will focus on the research of text detection,text location and binaration technology.Firstly, the paper comes to a conclusion that different algorithms have their advantage and disadvantage when finish the research and analysis about text detection, location and text image binaration, then present a text extraction approach based on gray and edge, it makes full use of the features that text has strong edge information and color contrast, in the processing period, morphological processing method and binaration method are used to delete the noise edge information. Lastly, present an improved binaration algorithm. In the last experiments, we apply the algorithms to news video retrieval system, the result shows that the algorithm run quickly and meet the requirements of real-time video retrieval.The paper have done a great amount of comparation between algorithm, and the experiment result shows whether it is adaptive. At the end of the paper, the paper points out disadvantages to be improved and the future research work about video text extraction.