What techniques are used to colorise old black and white movies? Are there different techniques? How to choose the right technique?
Convert movie to pictures (1 second = 24 pictures ; so 1 hour = 86,400 pictures ; average 90 min. movie = 129,600 pictures). Colorize every individual picture using the DeepLearning API which gives decent results comparable to the hand-colorized movies of the late 20th century. The Solomon Society uses this to convert old newsreel footage and other curiosities with impressive results.
It sounds like a lot of work, and it is (actually the work is simple but repetitive & massive), but a modern PC with an upper class video card, maximal RAM and a ton of free disk space should be able to pull it off fine given some time.
It is a largely manual and labor intensive process. Each object is manually identified and assigned colors. Computers can help by providing tracking of objects within a scene to apply some initial values for each frame, but it still comes down to manually picking the color of everything.
The problem is that black and white footage only provides information about luminance and luminance on an object is highly variable as it moves through light and shadow. There is no information at all to be able to determine the color, just how brightly the color would have been. The colors have to be chosen based on what people would expect it to be (green grass, skin tones, blue sky, etc).
There isn't any software available to consumers for doing this that I can think of. Might be able to find a plug in for After Effects or something, to make tracking masks and apply tints that way, but even that would be tricky. There isn't really an easy way to do it since it effectively comes down to painting the old footage, even if a computer provides some basic assistance.
Does Convolutional Neural Network based Image Colorization using OpenCV make sense to you? You may try it from Git colorization