摘要 |
PURPOSE:To perform video retrieval fast like text retrieval by performing collation based upon a character string derived from the feature quantity of video for the video retrieval. CONSTITUTION:The video to be retrieved is sequentially inputted, frame by frame, and temporarily stored in a memory. A representative frame extraction part GO detects a scene change from a frame image 27 in the memory by referring to the frame images before and after it and extracts a frame image 28 at the time of the detection. A character string conversion part 21 converts the frame image 28 into a character code string 29 on which its feature is reflected. Then a video name generation part 22 stores a constant-length shift register with the character code string sent from the character string conversion part 21. The character code string stored therein corresponds to the frame image string constituting the video and is outputted as a video name 30. A user, on the other hand, registers video for inquiry, indicates the start of the retrieval, and confirms the retrieval result at an interactive processing part 23.
|