A Rectification Method for Removing Rolling Shutter Distortion
We propose a rolling shutter video rectification method that can deal with both camera translation and rotation for videos obtained from unknown sources. As the exact distortions caused by rolling shutter are too complex, this method aims to remove the ma
- PDF / 504,033 Bytes
- 11 Pages / 439.37 x 666.142 pts Page_size
- 8 Downloads / 172 Views
A Rectification Method for Removing Rolling Shutter Distortion Gang Liu, Yufen Sun and Xianqiao Chen
Abstract We propose a rolling shutter video rectification method that can deal with both camera translation and rotation for videos obtained from unknown sources. As the exact distortions caused by rolling shutter are too complex, this method aims to remove the majority of the distortions. A 2D rotation model is used to approximately represent the motion of each frame. The parameters of this model are solved by minimizing the measurement constraints on point correspondences. To relax the restriction on the form of frame motion, the frame motion computation is performed sequentially. Experiments show that our method is comparable to the 3D models that require camera calibration.
Keywords CMOS image sensor Rolling shutter distortion Image rectification Post-processing technique
25.1 Introduction Nowadays, CMOS sensors are widely used in video cameras because of their low prices and the ability to integrate with image processing circuits. A CMOS camera commonly adopts rolling shutter to read and reset the rows of the sensor array sequentially, making the capture time of each row slightly after the capture time of the previous row. When camera or object moves during image capture, the scene imaged by each row presents relative motion, resulting in geometric distortion
G. Liu (&) Y. Sun X. Chen Intelligent Transportation System Research Center, Wuhan University of Technology, 430063 Wuhan, China e-mail: [email protected]
W. Lu et al. (eds.), Proceedings of the 2012 International Conference on Information Technology and Software Engineering, Lecture Notes in Electrical Engineering 210, DOI: 10.1007/978-3-642-34528-9_25, Ó Springer-Verlag Berlin Heidelberg 2013
235
236
G. Liu et al.
called rolling shutter distortion. These distortions degrade the quality of images and making objects in videos appear dynamic non-rigid deformation. Removing rolling shutter distortion needs to compute the relative displacements between rows in a frame. This computational task is more difficult than video stabilization in which only the relative motions between frames are computed. Recently the problem of removing rolling shutter distortions from videos has been studied by many researchers [1–8]. These researches focused on the distortions caused only by camera motion. 2D image processing methods were proposed to process the distortions induced by smoothly varying translation parallel to the image plane [2, 3, 6, 7]. The more complex distortions caused by camera rotation was studied using 3D rotation models [4], with a requirement of camera calibration. In this paper, we propose a rectification method that can handle both camera translation and camera rotation, without requiring any information about the camera. Since the general image transformation is very complex, we do not intent to estimate the motion of each pixel exactly, but only to compute approximate frame motions that can be used to remove the dominating di
Data Loading...
