A biologically inspired method for estimating 2D high-speed translational motion

Authors:
Baiqing Guan;Shigang Wang;Guobao Wang
Affiliations:
Institute of Mechatronics and Automation Technology, School of Mechanical Engineering, Shanghai Jiaotong University, 1954 Huashan Road, Shanghai 200030, PR China;Institute of Mechatronics and Automation Technology, School of Mechanical Engineering, Shanghai Jiaotong University, 1954 Huashan Road, Shanghai 200030, PR China;Institute of Mechatronics and Automation Technology, School of Mechanical Engineering, Shanghai Jiaotong University, 1954 Huashan Road, Shanghai 200030, PR China
Venue:
Pattern Recognition Letters
Year:
2005

Citing 6
Cited 1

Performance of optical flow techniques

International Journal of Computer Vision
Image Motion Estimation From Motion Smear-A New Computational Model

IEEE Transactions on Pattern Analysis and Machine Intelligence
Degraded Image Analysis: An Invariant Approach

IEEE Transactions on Pattern Analysis and Machine Intelligence
Estimation of motion parameters from blurred images

Pattern Recognition Letters
Estimating Piecewise-Smooth Optical Flow with Global Matching and Graduated Optimization

IEEE Transactions on Pattern Analysis and Machine Intelligence
Optical flow using textures

Pattern Recognition Letters

Measurement of sinusoidal vibration from motion blurred images

Pattern Recognition Letters

Quantified Score

Hi-index	0.10

Visualization

Abstract

Many industrial applications involve high-speed translational motions. Classical translation analysis methods like region-matching break down when large motion velocity causes motion blur in the image. In this paper, we rely on geometric moments to realize the spatial integration which may exist in the ''motion-from-blur'' mechanism of the biological vision, derive a theorem for the relationship between the geometric moments of the motion blurred image and the translational motion, and develop a novel motion-from-blur method based on this theorem for estimating 2D high-speed translation from motion blurred images. This approach utilizes the ''motion blur'' cue rather than neglect it, and therefore can compute the velocity and the acceleration of the motion from only two successive frames of motion blurred images while existing approaches estimate an accelerated motion from at least three images. Furthermore, this algorithm decomposes the 2D translation analysis problem into two 1D parallel problems to improve the real-time performance. Experimental results with both uniform motion and accelerated motion show that this method achieves good accuracy and efficiency.