Please don't expect wonders. If your video is as you describe then VE should provide better results than Lanczos, however when you upsize just 2 times the difference may be not stunning. It's becoming much more obvious when one upsizes 4 times or more.
how many consecutive frames does SR (quality mode) utilise to create the final the output frame?
In some sense, all of them. It processes frames one by one, accumulating information. On each frame it mixes spatially interpolated current frame N and result of work with the previous frame (high-res frame N-1) which in turn uses information from frame N-2 which uses info from frame N-3 and so on.
Oh ok, I'll try a 4x SR upsize to see the difference.
Also could my video compression options have anything to do with the result?
I am using Xvid single pass with the target quantizer set to 3 (Which is apparently the sweet spot for the filesize/quality tradeoff)
If you are saying that only the previous frame is used (which itself uses the frame before and so on).
Should'nt the quality of later frames be significantly better than earlier frames from the same scene? This gradual increase in quality would be a bit distracting so I assume your SR algorithim must include information from later frames as well as preceding frames
to create the final output frame.
I mean: say you take the frame no. 1000 of the final output video, which adjacent frames from the source video have been used to create this frame? eg. frames 990-1010 etc.
Is the number constant or is it dynamic and varies depending on scene/frame content etc. What are the differences between the Quality and speed SR modes?
