Research On Algorithm Of Extraction Of Caption In Television Video

With the advancement of television digitalization and IP (Internet Protocol) trends, the security of digital television signals for broadcasting and transmission becomes more serious. It become a hotspot of researching that how to effectively supervise the content of television and ensure the security of the content of video. The text information strongly relate to semantic of the television video. To effectively supervise the content of the video, we need analyze the text extracted from the television video. But due to the little size of the caption in the video and the text embedded in the complex background, it can not be recognized by the OCR (Optical Character Recognition) technique directly.We studied tow new technology related to caption extraction to resolve the problems in this paper:One is an algorithm based on corner points detection for caption detection. This algorithm detects the caption by corners for the density of the corners in the caption area is much bigger than others. We detect the corners in the video frames, and then detect the corners from the back of the image by clustering or other methods. The experiment results demonstrate the superiority of the proposed algorithm.The other one is for image binary using both global threshold and local threshold. In this algorithm, we binary the image by global threshold coarsely and then using local threshold to binary the processed image. And this algorithm holds the advantages of both global threshold and local threshold algorithm. It can restrain the noise as the global threshold algorithm, and also can keep the detail of the image after binarization as the part threshold algorithm.Besides, we build a video caption extraction system based on the algorithms mentioned .