Is matlab's optical flow object dense or sparse?

In the description of the vision.OpticalFlow object it states:

'The OpticalFlow System object™ estimates object velocities from one image or video frame to another. It uses either the Horn-Schunck or the Lucas-Kanade method.'

Im wondering if this computes the optical flow for each pixel in the frames (i.e. dense) or does it use sparse pixels (like a grid of random pixels throughout the frame)

Thank you!

Answer 1

[Dima Lisin]

It computes dense optical flow for every pixel. If you have the latest version (R2015b), please try the new optical flow classes: opticalFlowFarneback, opticalFlowHS, opticalFlowLK, and opticalFlowLKDoG.

For sparse optical flow, see vision.PointTracker.