Structural Videotext Regions Completion with Temporal-Spatial Consistency

With the rapid speed of spreading video content, some videotext exists in the video content. Usually some videotext is the advertisements and it is not needed for some people. Most video completion methods deal with the object removing or broken regions occurring when the object is undesired or original video is broken. But few methods can handle the completion of the original region occupied by motion videotext well since the complex background of motion videotext. Most existing algorithm took a lot computation time for video completion. However, the completed region should be generated as soon as possible based on the real-time consideration. In this paper, we propose a fast and regular structural videotext region completion method to recover the original structural region occupied by the scrolling videotexts. We utilize the characteristic of the canny edge map of frames to form structural edge in the occupied region with temporal consistency. The spatial texture region completion is carried out by our proposed video gradient interpolation approach. We have completed the whole system and the experimental results show that all of the horizontal and vertical scrolling videotexts can be completed well.