Title :
A Robust and Discontinuity-Preserving Approach to Optical Flow Estimation
Author :
Chen, Faling ; Luo, Haibo
Author_Institution :
Shenyang Inst. of Autom., Chinese Acad. of Sci., Shenyang, China
Abstract :
In this paper we develop a robust and discontinuity-preserving approach based on local and global combined model for the optical flow estimation. Firstly, the local and global optical flow methods are combined to yield dense flow fields that are robust against noise. Then a non-linear diffusion is provided to the model for discontinuity-preserving of optical flow boundary. Simultaneously, a coarse-to-fine multi-resolution strategy is presented for overcoming the difficulty of large motion estimation and the acceleration of convergence. Finally experiments with different sequences are performed demonstrating the proposed method´s excellent results.
Keywords :
image resolution; image sequences; motion estimation; coarse-to-fine multiresolution strategy; discontinuity-preserving approach; motion estimation; nonlinear diffusion; optical flow estimation; Acceleration; Automation; Convergence; Equations; Image motion analysis; Image sequences; Motion estimation; Noise robustness; Nonlinear optics; Optical noise;
Conference_Titel :
Image and Signal Processing, 2009. CISP '09. 2nd International Congress on
Conference_Location :
Tianjin
Print_ISBN :
978-1-4244-4129-7
Electronic_ISBN :
978-1-4244-4131-0
DOI :
10.1109/CISP.2009.5304725