Multi-image gradient-based algorithms for motion measurement using wavelet transform

Abstract

A multi-image wavelet transform motion estimation algorithm based on gradient methods is presented by using the characteristic of wavelet transform. In this algorithm, the accuracy can be improved greatly using data in many images to measure motions between two images. In combination with the reliability measure for constraints function, the reliable data constraints of the images were decomposed with multi-level simultaneous wavelet transform rather than the traditional coarse-to-fine approach. Compared with conventional methods, this motion measurement algorithm based on multi-level simultaneous wavelet transform avoids propagating errors between the decomposed levels. Experimental simulations show that the implementation of this algorithm is simple, and the measurement accuracy is improved.

References

1. Wang X K Broadmanufacturing theoryChinese Journal ofMechanical Engineering 2003 39(10)8694(in Chinese) 2. Sun J F Wang C Q Tracking method for movingvehicles based on feature-point optical flow and Kalman filtrationJournal of South China University of Technology (NaturalScience Edition) 2005 33(10)1923(in Chinese) 3. N Gupta L Kanal Gradient based image motionestimation without computing gradientsInternationalJournal of Computer Vision 1997 22(1)81101. doi:10.1023/A:1007931911114 4. Davis C Q Freeman D M Statistics of subpixel registrationalgorithms based on spatiotemporal gradients or block matchingOptical Engineering 1998 37(4)12901298. doi:10.1117/1.601966 5. Timoner S J Freeman D M Multi-image gradient-basedalgorithms for motion estimationOpticalEngineering 2001 40(9)20032016. doi:10.1117/1.1391495 6. Horn B K P Schunck B G Determining optical flowArtificial Intelligence 1981 17185203. doi:10.1016/0004‐3702(81)90024‐2 7. Horn B K P Weldon Jr E J Direct methods for recoveringmotionInternational Journal of ComputerVision 1988 25176. doi:10.1007/BF00836281 8. Bernard C P Discretewavelet analysis: a new framework for fast optic flow computationIn: Proceedings of the 5th European Conference onComputer VisionBerlin, HeidelbergSpringer-Verlag 1998 354368 9. Srinivasan S Chellappa R Optical flow using overlappedbasis functions for solving global motion problemsIn: Proceedings of the 5th European Conference on Computer VisionBerlin, HeidelbergSpringer-Verlag 1998 288304 10. Y T Wu T Kanade C C Li et al.Image registration using wavelet-based motion modelInternational Journal of Computer Vision 2000 38(2)129152. doi:10.1023/A:1008101718719 11. Chen L F Liao H Y M Wavelet-based optical flow estimationIEEE Transactions on Circuits and Systems for Video Technology 2002 12(1)112. doi:10.1109/76.981841 12. Mendelsohn J Simoncelli E Bajcsy R Discrete-time rigidity-constrained optical flowIn: Proceedings of the 7th International Conferenceon Computer Analysis of Images and PatternsBerlin, HeidelbergSpringer-Verlag 1997 255262 13. Liu H Y R Chellappa A Rosenfeld Fast two-frame multiscale dense optical flow estimationusing discrete wavelet filtersJournal ofthe Optical Society of America 2003 20(8)15051515. doi:10.1364/JOSAA.20.001505 14. Daubechies I TenLectures on Wavelets (in Chinese, trans. Li Jianping)BeijingNational Defense IndustryPress 2004 15. Mallat S G Atheory for multiresolution signal decomposition: the wavelet representationIEEE Transactions on Pattern Analysis and MachineIntelligence 1989 11(7)674693. doi:10.1109/34.192463 16. Zelnik-Manor L Irani M Multi-frame estimation of planarmotionIEEE Transactions on Pattern Analysisand Machine Intelligence 2000 22(10)11051116. doi:10.1109/34.879791 17. Lai S H B C Vemuri Reliable and efficient computationof optical flowInternational Journal ofComputer Vision 1998 29(2)87105. doi:10.1023/A:1008005509994 18. M Freeman D Aranyosi A J Gordon M J et al.Multidimensional motion analysis using computer microvision2005-11-18