The docs for OpticalFlowPyrLK don't explicitly state whether the pPrev array of initial coordinates of the feature-points is an input parameter or not. My impression however is that pPrev must be specified by the caller. The IPPI demo code in OpticalFlowPyrLK.c on the other hand, simply sets the pPrev array to zero values.
My questions:
1. So, is this just bad sample code or an undocumented something where OpticalFlowPyrLK figures out good feature points itself ?
2. OpticalFlowPyrLK is based on "Pyramidal Implementation of the Lucas Kanade Feature Tracker" by Jean-Yves Bouguet, of the Intel Corporation Microprocessor Research Labs. Chapter 3 "Feature Selection" of this document describes an the algorithm to select feature-points. How does that algorithm translate into IPPI calls ?
Regards,
Adriaan van Os