| 摘要 |
Provided is a video search device with which a user readily specifies the position, orientation, or the like of an object expected to appear in a video, and the device searches the video therefor. In the video search device (501), a receiver (502) receives an input of a static image, 2 reference positions in the static image, and 2 target positions in a video frame. An extracting unit (503) extracts a reference image including the 2 reference positions from the static image. A searching unit (504) searches the frame images included in the video for a similar frame image depicting a local image similar to the reference image, tracks 2 movement loci when time is advanced or reversed from the similar frame image in the video, said 2 movement loci being associated with 2 pixels of interest depicted in starting positions corresponding to the 2 reference positions, searches for a target frame image in which the 2 movement loci are in the vicinity of the 2 target positions, and sets video including the similar frame image and the target frame image as the search result. |
