I don't see how GPUs are relevant. The question wasn't about real time video, but video without flickering. I don't have the domain knowledge to answer confidently, but the algorithm was designed to be noise-tolerant. From my own amateur-level reading of the paper I see no reason why it shouldn't work well with video.
Thanks; worked like a charm once I pulled in some more libraries. One thing to tweak for Linux use: add upper-case versions of the image file extensions to the fifth parameter of the openFileDialog call in gui.cpp, so you don't have to rename all those image files that cameras and phones tend to give upper case names to.
I wonder if given the input and output of the algorithm, you could train some type of deep neural net. And if successful, what the hidden layers would look like?
Maybe I'm missing something, but frankly I think if you're building GUI with Python, tkinter is probably the best option. Comes with stock, is pretty solid, and on edge-cases serves you well.
I had to create some applications with py earlier this year, I started out with pyside but it was really heavy and kind of difficult to grok. wxwidgets I tried as well and it didn't work (I needed mouse cursor information and use it in a certain way to draw certain things, it was messy to do it with everything except tkinter).
The comparison with more basic edge detect algorithms, at the end, is interesting.