1
0
mirror of https://github.com/opencv/opencv_contrib.git synced 2025-10-18 00:01:17 +08:00

Added a note about key differences

This commit is contained in:
Vladislav Samsonov
2016-07-08 22:20:03 +03:00
parent 1d90d67016
commit 3dbbad123a

View File

@@ -48,6 +48,12 @@ http://files.is.tue.mpg.de/black/papers/cvpr2015_pcaflow.pdf
month = jun,
year = {2015}
}
There are some key differences which distinguish this algorithm from the original PCAFlow (see paper):
- Discrete Cosine Transform basis is used instead of basis extracted with PCA.
Reasoning: DCT basis has comparable performance and it doesn't require additional storage space.
Also, this decision helps to avoid overloading the algorithm with a lot of external input.
- Usage of built-in OpenCV feature tracking instead of libviso.
*/
#ifndef __OPENCV_OPTFLOW_PCAFLOW_HPP__