| 摘要 |
In order to retrieve text data from a video signal comprising a series of frames (such as a broadcast television signal, or a playback signal of a recording of a broadcast television signal), a sequence of the frames from the video signal is captured. For each captured frame in the sequence, it is determined whether or not a ratio of luminance of a brightest part of that captured frame relative to average luminance of that captured frame exceeds a ratio threshold. If so, a character recognition process is performed on that frame to detect text in that frame. The invention takes advantage of the fact that, in order for the text to be legible to the viewer, it is generally enhanced relative to the background by giving it much greater luminance than the rest of the image. A simple luminance ratio test on the frame can therefore be used to identify frames likely to contain text, and the character recognition process need be performed only on those identified frames. Typically, the number of frames containing text is far less than the total number of frames in a television programme, and so this opens up the possibility of retrieving text data in "effective" real time. <IMAGE> <IMAGE> <IMAGE> |
