next up previous
Next: References Up: 3-D Surface Normals and Previous: Basket Sequence

Conclusions

This work has demonstrated that affine flow parameters can be used to recover 3-D structure and motion, giving denser structure description via the surface normals than available from a feature based approach, whilst avoiding the high computational demands of an optical flow scheme. The method performed well when the affine model provided a good approximation of the local motion fields (particularly for near-view textured surfaces), even when the measurements contained significant amounts of noise. Moreover, its ability to recover stable estimates of the focal length was encouraging and appears to provide a robust method for obtaining metric estimates without the need for pre-calibration.



Andrew Calway
Mon Dec 4 11:27:23 GMT 2000