Title :
Robust optical flow for Driver Assistance
Author :
Xu, Jie ; Chen, Hui ; Ding, Weifu ; Zhao, Changsheng ; Morris, John
Author_Institution :
Sch. of Inf. Sci. & Eng., Shandong Univ., Jinan, China
Abstract :
Optical flow in Driver Assistance Systems (DAS) must be robust to variations in illumination and handle large disparities in real time. Although a pyramidal variation of Total Variation Regularized L1 (TV-L1) optical flow can handle large disparities, TV-L1 is not adequate for real world DAS since the ambient light changes dynamically in typically encountered scenes. We present a novel improved TV-L1 with high boost filtering, which by enhancing high frequency components, becomes robust to illumination changes. Experiments show that our method can measure large disparities even under large dynamic illumination variations.
Keywords :
driver information systems; image sequences; DAS; TV-L1; Total Variation Regularized L1; driver assistance systems; frequency components; pyramidal variation; robust optical flow; Adaptive optics; Cameras; Lighting; Optical imaging; Optical sensors; Robustness; Vehicles; DAS; TV-L1; high boost filtering; optical flow;
Conference_Titel :
Image and Vision Computing New Zealand (IVCNZ), 2010 25th International Conference of
Conference_Location :
Queenstown
Print_ISBN :
978-1-4244-9629-7
DOI :
10.1109/IVCNZ.2010.6148834