The present invention is discloses a video text recognition system. The system includes a fusion concept module receiving a raw video incorporated with text, improve the quality of low resolution video based on neighbor information of pixels in temporal information, a Ring Radius Transform (RRT) detects the text line from video, a text localization module separating text from background by using a bit plane concept, a semantic feature module establishing a spatial relationship between the pixels in different domain, such as gradient, color, Laplacian in new way; and a recognition module matches with a template stored in a template matching module.
Komen