DocumentCode :
3404781
Title :
Fast automatic elimination of vertical parallax of multiview images
Author :
He, Wang ; Guozhong, Wang ; Liliang, Liu ; Yang, Zhang ; Ping, An ; Zhaoyang, Zhang
Author_Institution :
Dept. of Telecommun., Shanghai Univ., Shanghai, China
fYear :
2010
fDate :
24-28 Oct. 2010
Firstpage :
1004
Lastpage :
1007
Abstract :
A fast and automatic method is proposed to eliminate the vertical parallax in multiview images. The horizontal parallax provide the viewers stereoscopic feeling, but the vertical parallax causes the view discomfort in autostereoscopic display. To decrease the viewing discomfort, the elimination of vertical parallax is necessary. The complexity of traditional processing method is discussed, and the complexity of the current methods is very high. Thus it is difficult to do real time rectification of 3DTV video images. In the proposal, SIFT(Scale Invariant Feature Transform) is employed to detect the features among different views. And a compact processing procedure is proposed to lessen the complexity of traditional methods automatically without complicated parameters initialization. The experiment shows that the performance of the proposal is encouraging as well as the speed of the processing is fast. Thus, it is convenient to implement for real time 3D video acquisition and 3DTV displaying.
Keywords :
feature extraction; stereo image processing; television displays; three-dimensional television; 3D video acquisition; 3DTV displaying; 3DTV video images; autostereoscopic display; fast automatic elimination; horizontal parallax; multiview images; scale invariant feature transform; vertical parallax; Cameras; Complexity theory; Feature extraction; Proposals; Real time systems; Streaming media; Three dimensional displays; 3DTV; automatic; fast; multiview; vertical parallax;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Signal Processing (ICSP), 2010 IEEE 10th International Conference on
Conference_Location :
Beijing
Print_ISBN :
978-1-4244-5897-4
Type :
conf
DOI :
10.1109/ICOSP.2010.5655852
Filename :
5655852
Link To Document :
بازگشت