Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

Lucas kanade optic flow Winsize parameter


I am trying to calculate optic flow from an image using the opencv cpp implementation of the pyramidal Lucas-Kanade algorithm: cv.calcOpticalFlowPyrLK()

I am confused as for what the winSize parameter stands for. It is defined as: size of the search window at each pyramid level. Does that mean it refers to the neighborhood of points around the central pixel p you assume to optic flow to be "similar" in order to solve the optical flow equation?

The opencv documentation says:

We have seen an assumption before, that all the neighbouring pixels will have similar motion. Lucas-Kanade method takes a 3x3 patch around the point. So all the 9 points have the same motion.

Is this 3x3 the winSize parameter? If so, why is there an option to choose the parameter If it is fixed to 3?