How to compute optical flow using tvl1 opencv function

Question

I'm trying to find python example for computing optical flow with tvl1 opencv function createOptFlow_DualTVL1 but it seems that there isn't enough documentation for it.

Could anyone please let me do that? I've used calcOpticalFlowFarneback mentioned here http://docs.opencv.org/master/d7/d8b/tutorial_py_lucas_kanade.html but it is not giving me accurate results, will tvl1 good enough and if not is there another method I should look for?

[[EDIT]]

I've some regions come from selective search, I want keep only regions with motion in it, so computing the OF for a given frame and then get the avg in each region could do it. It's also described in this paper section 3.1

Thanks.

Answer 1

Change this line(Dense Optical example in http://docs.opencv.org/master/d7/d8b/tutorial_py_lucas_kanade.html):

flow = cv2.calcOpticalFlowFarneback(prvs, next, None, 0.5, 3, 15, 3, 5, 1.2, 0)

By these:

optical_flow = cv2.DualTVL1OpticalFlow_create()
flow = optical_flow.calc(prvs, next, None)

The parameter descriptions can be found here: http://docs.opencv.org/3.3.0/dc/d47/classcv_1_1DualTVL1OpticalFlow.html