Abstract
A multi-scale optical flow algorithm was proposed, which uses oriented smoothness constraint to propagate and optimize the local and rough estimation of the optical flow. Under the frame of multi-scaled Laplacian pyramid, the sum-of-squared difference (SSD) matching technique estimates the flow velocity and the covariance matrix in searching area. Then the oriented smoothness constraint is employed to propagate and optimize the flow velocity estimated, which can suppress the over-smoothness across the occlusions area and motion discontinuities. The experiments obtained encouraging results using synthetic and real image sequences.
| Original language | English |
|---|---|
| Pages (from-to) | 153-158 |
| Number of pages | 6 |
| Journal | Gaojishu Tongxin/High Technology Letters |
| Volume | 17 |
| Issue number | 2 |
| Publication status | Published - Feb 2007 |
| Externally published | Yes |
Keywords
- Laplacian pyramid
- Multi-scale
- Optical flow
- Oriented smoothness constraint
- Sum-of-squared difference (SSD)