Python implementation of optical flow estimation using only the Scipy stack for:

• Horn Schunck

Lucas-Kanade is also possible in the future, let us know if you're interested in Lucas Kanade.

python -m pip install -e .

optionally, to run self-tests:

python -m pip install -e .[tests]

pytest -v

The program scripts expect directory glob pattern

imageio loads a wide varity of images and video.

Box:

python HornSchunck.py src/pyoptflow/data/tests/box box*.bmp

Office: all time steps:

python HornSchunck.py src/pyoptflow/data/tests/office office*.bmp

or just the first 2 time steps:

python HornSchunck.py src/pyoptflow/data/tests/office office.[0-2].bmp

Rubic:

python HornSchunck.py src/pyoptflow/data/tests/rubic rubic*.bmp

Sphere

python HornSchunck.py src/pyoptflow/data/tests/sphere sphere*.bmp

Compare: Matlab Computer Vision toolbox: in matlab, similar method in Octave and a comparison plot using Matlab Computer Vision toolbox.

Reference:Inspiration