Why we need a coarse-to-fine strategy to solve the optical flow problem (feature tracking) in practice? If we do not use such kind of methods, what will happen?
Asked
Active
Viewed 387 times
1 Answers
0
If an image has larger motions, then the traditional Optical Flow methods like the non-iterative Lucas Kanade can perform badly because they assume motion is small (subpixel). The coarse-to-fine method helps a lot because in a coarser resolution you can assume that the brightness constancy can hold.
insanely_sin
- 986
- 1
- 14
- 22