DocumentCode :
3003313
Title :
Telop character extraction from video data
Author :
Kuwano, Hidetaka ; Kurakake, Shoji ; Odaka, Kazumi
Author_Institution :
NTT Human Interface Labs., Kanagawa, Japan
fYear :
1997
fDate :
35601
Firstpage :
82
Lastpage :
88
Abstract :
We introduce a telop character extraction method that consists of two steps: detecting a telop frame from a video sequence; and extracting telop character regions in the telop frame. Frames in which telops appear and disappear are detected regardless of background movement in the video sequence by selecting frames that have large intensity histogram differences from the previous frames and have large agreement in edge pixel positions with the subsequent frames. By segmenting the color space in a sequential manner, the method then quickly segments the detected telop frames without deteriorating segmentation results. It then filters out non-telop character regions obtained from the segmentation by using color, location, size and temporal movement features of the regions. We implement the proposed method on a PC and a workstation, and confirm that it can process live broadcast video data very effectively
Keywords :
feature extraction; image colour analysis; image matching; image segmentation; image sequences; microcomputer applications; optical character recognition; character regions; color space segmentation; edge pixel positions; image color; large intensity histogram difference; live broadcast video data; location; personal computer; size; telop character extraction; telop frame detection; temporal movement features; video data; video sequence; workstation; Content based retrieval; Data mining; Filters; Histograms; Humans; Image edge detection; Information retrieval; Laboratories; Multimedia communication; Video sequences;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Document Image Analysis, 1997. (DIA '97) Proceedings., Workshop on
Conference_Location :
San Juan
Print_ISBN :
0-8186-8055-5
Type :
conf
DOI :
10.1109/DIA.1997.627096
Filename :
627096
Link To Document :
بازگشت