SYSTEM FOR AUTOMATIC VIDEO SEGMENTATION AND KEY FRAME EXTRACTION

摘要

PURPOSE: To provide a highly reliable contents analysis system by distinguishing a sharp break generated by specific action from a gradual transition, and automatically and temporarily dividing a video sequence to individual camera shorts. CONSTITUTION: After initializing a reference system parameter in a block 101 at first, a 1st reference frame is selected in a block 102 and a difference D1 between frames is calculated based on a difference measuring reference selected in a block 104. When the D1 is larger than a previously set threshold, a change in a shot is detected and recorded in a block 106 prior to advance to a block 108. In other cases, the video sequence is directly advanced to the block 108 to form a new frame. When the frame concerned is a final frame, a segmentation process is ended, but if not, the video sequence is advanced to a block 103 and the process is repeated until reaching the final frame of the video sequence.