trade motion vectors for spatial resolution, 60FPS-->30FPS
I have some captured video AVC1 1080p 60FPS that is only 8Mbps (too low). For portions with little motion, it looks good. If there's motion, the detail becomes blurred. I suspect if they had done 1080p 30FPS 8Mbps or 1080p 60FPS 16Mbps it would look fine regardless of motion.
I'm thinking that with the right algorithm, it may be possible to gain detail (spatial resolution) by using temporal motion from the discarded frames. In other words, going from 60FPS to 30FPS, there may be a way to interpolate additional detail via the motion vectors.
With Avisynth and the right filters and settings, it could calculate motion vectors for each frame and save them to a temp file, toss every other frame, then use those calculated vectors to produce a video with enhanced detail but at half the FPS.
Code and comments would be much appreciated.
|