摘要 |
PROBLEM TO BE SOLVED: To create a label conforming to the content of a scene in a video content to present the scene to a user so that the user can easily understand the content of the scene. SOLUTION: Class scores SC1 to SC6 of respective classes which show the content of an image are calculated as frame scores FSC by using image feature quantities FSP of a frame image F3. Based on the frame scores FSC to a shot delimiting point, a class average value of the class scores of the individual classes is calculated as a shot score SHC of a shot SHT. Then, a shot label for the shot SHT is created based on a single or a plurality of shot scores SHC. Scene delimiting points of a scene are detected based on the shot labels of the plurality of shots SHT, and at the same time, a scene label can be created based on the plurality of shot labels, and hence a scene label conforming the content of a scene in a video content can be more accurately created. COPYRIGHT: (C)2011,JPO&INPIT
|