Abstract
In this paper we propose a novel technique for estimating motion of two-dimensional objects. The technique involves representing objects uniquely by straight-line approximations of the boundary, {(θ, p)} and estimating motion from shifts in the θ - p space. The straight-line approximation of the boundary is obtained by detecting dominant points during contour tracing and then approximating the pixels between adjacent dominant points (along the contour) by a straight line using the Hough transform. We have developed a computationally efficient way of detecting dominant points in a single pass during contour tracing. The proposed object representation enables us to determine the translation and rotation with significantly fewer computations compared with existing methods. Our method can also detect the motion parameters correctly for partially overlapping objects.
Original language | English (US) |
---|---|
Pages (from-to) | 1245-1258 |
Number of pages | 14 |
Journal | Pattern Recognition |
Volume | 29 |
Issue number | 8 |
DOIs | |
State | Published - Aug 1996 |
Keywords
- Contour tracing
- Dominant point
- Hough transform
- Inverse hough transform
- Motion estimation
- Overlapping objects
- Region of support
ASJC Scopus subject areas
- Software
- Signal Processing
- Computer Vision and Pattern Recognition
- Artificial Intelligence