Video Caption Recognition Research

Along with the rapid development in computer, multimedia and networking technologies, digital video applications become more and more popular. How to fast and efficiently retrieve the extremely large amount of video data is the key problem. This paper does from caption and use caption as a clue for video retrieval. The caption-pictures which were Extracted from videos and transformed to pure text are saved as text formatting, then we use these caption-text to retrieve video and give an approach to video retrieval. The video caption recognition mainly contains steps of caption detection, caption location, caption extraction, caption enhancement and character recognition. This paper puts the caption detection, caption location, caption extraction and character recognition as the mainly content to research and take two steps such as caption-frame detection and caption of frame-image detection to detect caption.Firstly, the background and the development status of CBVR technology are introduced. Secondly, synopsizes and analyzes the traditional methods of caption extraction. Thirdly, Based on the traditional methods, theory and practice, a new method of caption detection is put forward; it well raises the efficiency and the rate of accuracy of the old methods. By anglicizing the changing of the colors of the background and caption-location of the frame-image, a new method of caption detection and caption location based on partition is put forward. By anglicizing the features of the RGB color space and the LAB color space, a new method of caption extraction based on LAB color space is put forward, and it reduces the color losing of using the binaryzation in the RGB color space. By anglicizing the features of the characters of video captions and the common characters, a new method of character recognition based on feature compensating which are mainly to the video caption character is put forward, getting a very well recognizing effection.