Approach Based on Blocking and Lab Color Model for Caption Extraction in Video Frames

Captions is an important clue for setting up video retrieval,because the captions are always relevant to the video frames and objectively descript the main contents of the video frames. In this paper,according to that the pixels's color of the background change slowly while the pixels's color of the junction of the background and foreground change quickly,using approach that blocking every row of the video frame and computing the changing rate of every blocks to realize caption detection and location. Using Lab color model to realize binary text extraction. Experiment results show this approach is promising and has a good application prospects.