Originally Posted by MKANET
I'm curious whats the visual difference between SVP framerate interpolation and adjustable framerate interpolation available on most >240hz LCD TVs besides the processing being done outside the PC? I get pretty nice results when I set framerate interpolation on my TV to 30%; which equals to about 72fps on 24fps source material. Frames are matched perfectly with vertical refresh rate since its all done within the TV. TV detects both 24.000fps using reclock (which is what I use) or 23.976fps.
The basics are the same, the difference is how each scheme treats fast moving scene changes and the resultant motion artifacts is an indication of the accuracy of that effort. I assume your TV doing proper interpolation and not frame insertion/deletion ala pull up/down schemes of the past.
SVP's motion estimation is vector adaptive and it uses multiframe lookahead to see how the scene changes much like a multipass H.264 encoder.
It can do this as it has enough memory in the PC to do so, most SoC based implementations used in TVs do not have the luxury.
Things may change when CE manufacturers can afford to dedicate 4-8Gb of memory for video processing.
The other difference is the amount of processing power that is available, in order to perform good motion estimation, the interpolator needs to search both forward and back in time and this is mathematically complex, and is why GPUs are so useful for this task.
In many respects the motion estimation engine in SVP is not too different from those used in multipass high end H.264 encoders, except it works in real time.Edited by Tong Chia - 12/27/12 at 1:39pm