|
|
|
Sub-pixel motion detection by joint transform correlator
based on bicubic spline interpolation |
| GE Peng, LI Qi, FENG Hua-jun, XU Zhi-hai, CHEN Yue-ting |
| State Key Lab of Modern Optical Instrumentation, Zhejiang University, Hangzhou 310027, China |
|
|
|
Abstract For joint transform correlator (JTC) can only detect motion vector in integral pixel, a method using bicubic spline interpolation to increase detection accuracy to sub-pixel is proposed. First the integral pixel motion vector between two images was detected by JTC, and then a 5×5 area around peak value coordinates were selected. A dualistic surface was acquired by using bicubic spline interpolation. Through differentiating the dualistic surface, motion vector was achieved. Then the pictures were recovered by fitting. Moreover, the effects of stabilization of sampling accuracy in 0.01 pixels and 0.1 pixels were contrasted. Experimental results show that the differences of PSNR(peak signal-noise rate) between them are not significant .But when sampling rate is 0.1 pixel, it satisfies the stabilization’s need for real-time. Motion detection by JTC based on bicubic spline interpolation rises to 0.1 pixels. PSNR has been increased．
|
|
Published: 23 December 2010
|
|
|
双三次样条插值联合变换相关器亚像素探测技术
针对传统光学联合变换相关器只能探测到整数像素的位移矢量,提出用双三次样条插值方法提高联合变换光相关器探测位移矢量精度到亚像素.用光学联合变换相关器探测2幅图像的整数像素位移,取峰值整数坐标的5×5邻域,用双三次样条插值对此邻域进行二元曲面拟合,对曲面求导得到亚像素位移矢量,再用拟合法来复原图像.对比采样精度分别为0.01、0.1像素时的稳像效果,结果表明，两者峰值信噪比相差不大,但采样精度为0.1像素时,所花时间更能满足稳像实时性的要求.基于双三次样条插值的光学联合变换相关器探测精度达到0.1像素,峰值信噪比得到了较大提高.
|
|
| ［1］ 汪小勇,李奇,徐之海,等.用于实时数字稳像的灰度投影算法研究［J］.光子学报,2006,35(8):1268-1271.
WANG Xiaoyong, LI Qi, XU Zhihai, et al. Realtime digital image stabilization system based on gray projection algorithm ［J］. Acta Photonica Sinica, 2006, 35(8):1268-1271.
［2］ 李奇,徐之海,冯华君,等.基于刚性运动模型的全局运动矢量探测方法［J］.测试技术学报,2006,20(6):524-528.
LI Qi, XU Zhihai, FENG Huajun, et al. Detection method for global motion vector based on rigid motion model ［J］. Journal of Test and Measurement Technology, 2006, 20(6):524-528.
［3］ 陈跃庭,李奇,许东晖,等. 投影算法与块匹配法结合的实时数字稳像方法［J］. 浙江大学学报:工学版,2007,41(9):1537-1540.
CHEN Yueting, LI Qi, XU Donghui, et al. Realtime digital image stabilization method based on projection algorithm and block matching ［J］. Journal of Zhejiang University: Engineering Science, 2007, 41(9):1537-1540.
［4］ JANACEK K, TCHERNYKH V, DYBLENKO S, et al. Integrated camera motion compensation by realtime image motion tracking and image deconvolution ［C］∥Proceedings of the 2005 IEEE/ASME International Conference on Advanced Intelligent Mechatronics Monterey.California:IEEE/ASME.2005, 1437-1444．
［5］ 陈文略,王子羊. 三次样条插值在工程拟合中的应用［J］.华中师范大学学报：自然科学版,2004,38(4):418-422.
CHEN Wenlue, WANG Ziyang.The application of cubic splines in terpolation in the project fitting ［J］. Journal of Central China Normal University:Nature Science,2004, 38(4):418-422.
［6］ CHAO T H, ZHOU H Y, REYES G.Compact 512×512 grayscale optical correlator［C］∥ Proceedings of SPIE.Orlando:FL,USA:\
[s.n.\], 2002, 4734(9):9-12.
［7］ ALAM M S, KARIM M A. Fringeadjusted joint transform correlation ［J］. Apply Optic, 1993,32(23): 4344-4350.
［8］ 王爱玲.计算几何及图形处理(CAD基础)［M］.北京：国防工业出版社,1988:1280．
［9］ MORIMOTO C, CHELLAPPA R. Evaluation of image stabilization algorithms［C］∥IEEE Proceedings of ICASSP.WA: IEEE, 1998: 2789-2792. |
|
Viewed |
|
|
|
Full text
|
|
|
|
|
Abstract
|
|
|
|
|
Cited |
|
|
|
|
| |
Shared |
|
|
|
|
| |
Discussed |
|
|
|
|
