Users Guide

To estimate the classical Optical Flow using the original technique proposed by Horn & Schunck, use the functor bob.ip.optflow.hornschunck.VanillaFlow. In the following example, we assume you provide a pair of gray-scaled images i1 and i2, represented using a 2D numpy.ndarray containing numpy.float64 elements:

 >>> import bob.ip.optflow.hornschunck
 >>> flow = bob.ip.optflow.hornschunck.VanillaFlow(i1.shape)
 >>> u, v = flow.estimate(alpha=200, iterations=20, image1=i1, image2=i2)
 >>> print(u)
 [[...]]
 >>> print(v)
 [[...]]

If you’d like to estimate the flow for an image, given a triplet, then use the bob.ip.optflow.hornschunck.Flow instead:

 >>> import bob.ip.optflow.hornschunck
 >>> flow  = bob.ip.optflow.hornschunck.Flow(i1.shape)
 >>> u, v = flow.estimate(200, 20, i1, i2, i3)
 >>> print(u)
 [[...]]
 >>> print(v)
 [[...]]