DocumentCode :
1616143
Title :
Gravity-oriented nonlinear stereo rectification for stereoscopic 3D
Author :
Zhao, Yu ; Shao, Hang ; Dai, Qionghai
Author_Institution :
Dept. of Autom., Tsinghua Univ., Beijing, China
fYear :
2011
Firstpage :
1
Lastpage :
4
Abstract :
We propose a novel and practical stereo rectification method. The method is both simple and efficient for 3D content creation, especially for 3D post-production. The major contribution of this paper is rectifying the stereo video with a nonlinear deformation considering the gravity orientation, which helps producing view-reasonable and non-gap content instantly. In our framework, a new projective rectification introducing gravity orientation into the process is firstly calculated, which better suits 3DTV production than conventional methods. We then apply a nonlinear deformation derived from earlier projective transformation, preventing from the unpleasant gap during the whole process. Experimental results have demonstrated that our method obtains better visual effect than the state-of-the-art methods. Additionally we apply CUDA parallel computation to solve nonlinear deformation, which is of great significance for the real-time application in post-production.
Keywords :
deformation; stereo image processing; television production; three-dimensional television; video signal processing; 3D content creation; 3D post-production; 3D stereoscopic; 3DTV production; CUDA parallel computation; gravity-oriented nonlinear stereo rectification method; nonlinear deformation; projective transformation; stereo video; Gravity; Image edge detection; Nonlinear distortion; Pixel; Real time systems; Shape; Three dimensional displays; 3D Video; gravity-oriented; nonlinear deformation; stereo rectification;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
3DTV Conference: The True Vision - Capture, Transmission and Display of 3D Video (3DTV-CON), 2011
Conference_Location :
Antalya
ISSN :
2161-2021
Print_ISBN :
978-1-61284-161-8
Type :
conf
DOI :
10.1109/3DTV.2011.5877160
Filename :
5877160
Link To Document :
بازگشت