Paper
12 March 2013 Accurate depth estimation using spatiotemporal consistency in arbitrary camera arrays
Woo-Seok Jang, Yo-Sung Ho
Author Affiliations +
Proceedings Volume 8648, Stereoscopic Displays and Applications XXIV; 86481K (2013) https://doi.org/10.1117/12.2004223
Event: IS&T/SPIE Electronic Imaging, 2013, Burlingame, California, United States
Abstract
Depth estimation is an essential task for natural 3D image generation. In this paper, we estimate an accurate depth map from stereoscopic images captured by arbitrary camera arrays. Usually the depth information is estimated by stereo matching from two input images that are obtained by the parallel camera array. Recently the arc camera array has been widely employed to produce 3D movies. However, in the convergent camera array, it is difficult to apply image rectification by matching correspondence points due to serious image distortion. In this work, we estimate depth data without using image rectification. Once we define a potential energy function for depth detection based on spatial consistency, the energy optimization process identifies mismatching depth pixels. A reasonable depth value is assigned to each mismatched pixel using distance and intensity differences between the mismatched pixel and its neighbors. In addition, we improve temporal consistency and reduce visual discomfort. Experimental results demonstrate that our proposed method provides more accurate depth values than other methods based on image rectification.
© (2013) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Woo-Seok Jang and Yo-Sung Ho "Accurate depth estimation using spatiotemporal consistency in arbitrary camera arrays", Proc. SPIE 8648, Stereoscopic Displays and Applications XXIV, 86481K (12 March 2013); https://doi.org/10.1117/12.2004223
Advertisement
Advertisement
RIGHTS & PERMISSIONS
Get copyright permission  Get copyright permission on Copyright Marketplace
KEYWORDS
Cameras

Image processing

3D image processing

3D acquisition

Video

Visualization

Distortion

Back to Top