Strange error calculation in calcOpticalFlowPyrLK (Bug #1651)
The documentation says that output error in calcOpticalFlowPyrLK is calculated as difference between patches around the original and moved points.
In iteration cycle point on the second image is used as center of patch when we compute difference between patches.
But in the end when we compute error, point on the second image is used as left top corner of patch.
Is this right? Or it should be the center of patch?