| 摘要 |
<p><P>PROBLEM TO BE SOLVED: To easily detect a caption text from a video including a program video and a caption video. <P>SOLUTION: A video caption detection device comprises: a geometric conversion unit 12 which takes in imaged video data, and detects a television screen domain from a frame picture of the imaged video data on the basis of a rectangle formed by linear line components detected from the frame picture; a sub screen estimation unit 13 which detects picture edges from the television screen domain, and estimates a sub screen domain from the frame picture on the basis of distribution of the edges; a caption domain estimation unit 14 which detects change in pixels in the time direction from the sub screen domain, and estimates a caption domain on the basis of the change in the pixels; a sub screen domain detection unit 17 which takes in the imaged video data, and detects the sub screen domain from the frame picture; and a caption information extraction unit 18 which detects the caption domain from the sub screen domain, extracts a caption text from the caption domain, generates caption information mapping the text and time information showing the current time, and outputs the information. <P>COPYRIGHT: (C)2013,JPO&INPIT</p> |
