I recently updated my hardware and came to the idea to speed up an application of mine that analyses movie files (say: compute the differences of adjacent frames) to do parallel. The idea was to assign the file partwise to the cores/ threads, to do reading/ processing indepently and after that do the stitching (necessary because of inaccuracies). I tried that and got very disappointing speed-ups. I posted that to stackoverflow as I don't know where the cause is (os, opencv, codecs, hardware, ...) but didn't get helpful feedback.
So my try specific in this forum. As it seems a task of general interest.
Is it possible to speedup reading in/ analysing a movie file using opencv? And how do it best? (My attempt seems to have same flaw).