Study On Caption Text Detection And Location Of News Video

Paper takes Content-Based Video Retrieval (CBVR) technology in news video edition system application as research background. It studies some key questions about news video Caption Text Detection and Location (CTDL), and it has the certain theory significance and the widespread application prospect.On the base of thorough research about news video caption text character, paper contrasts each kind of image edge detection algorithm to the news video serviceability. It analyzes and compares many kinds of image enhancement method including space filter enhancement and frequency filter enhancement and so on. In this foundation, paper proposes one kind video frame image enhancement method based on the Mathematical Morphology operators multiple operations. Every news video frame is carried on the gradation morphology closing at first, and then carried on the gradation morphology opening, finally carried on binary morphology erosion. The algorithm serviceability is confirmed through experiments.In the foundation of research on news video caption edge detection general method, paper improves caption text and background density distinctness algorithm. It studies with emphasis two class classifier edge detection and location methods based on two problems. The result improves detection precision, and the effect obviously surpasses the traditional projection.Taking advantage of information redundancy in news video, paper studies the sample frame caption text detection and location. It automatically tracks and matches the sample frame neighboring frame caption. The method not only reduces calculation times, but also effectively filters some false alarm.Based on above studies, author has built one system frame construction about from edge detection to text location in news video caption detection and location. The steadiness of algorithm in paper has been confirmed through experiment. Two papers have been published or accepted, where is Information Journal of Chinese Core Journal of information class and Computer Simulation of Science and Technology Core Journal of computer class.