Title :
Finite-differencing errors in gradient-based optical flow estimation
Author :
Brandt, Jonathan W.
Author_Institution :
Sch. of Inf. Sci, Japan Adv. Inst. of Sci. & Technol., Ishikawa, Japan
Abstract :
The effect of partial derivative approximation errors on gradient-based optical flow estimation is studied through frequency-domain analysis and simulations. It is found that the non-ideal response of a finite-differencing convolution kernel contributes significantly to optical flow estimation errors. The results suggest that it is necessary to weigh such factors as input spatial bandwidth and maximum velocity magnitude when selecting a finite-differencing kernel for gradient-based optical flow estimation
Keywords :
convolution; error analysis; finite difference methods; frequency-domain analysis; image sequences; motion estimation; finite-differencing convolution kernel; finite-differencing errors; finite-differencing kernel; frequency-domain analysis; gradient-based optical flow estimation; input spatial bandwidth; maximum velocity magnitude; nonideal response; partial derivative approximation errors; simulations; Analytical models; Approximation error; Bandwidth; Convolution; Estimation error; Finite difference methods; Frequency domain analysis; Frequency estimation; Image motion analysis; Kernel;
Conference_Titel :
Image Processing, 1994. Proceedings. ICIP-94., IEEE International Conference
Conference_Location :
Austin, TX
Print_ISBN :
0-8186-6952-7
DOI :
10.1109/ICIP.1994.413676
