Hi all,
I’ve taken a look at the video segment. The problem is pretty straightforward — That segment is constantly in motion. The motion isn’t fast, but the speed of the motion isn’t at all important. With constant motion, Peach Smoother can’t ever get a noise estimate. More importantly — With constant motion throughout the picture, you really shouldn’t use a temporal smoother at all.
Without a noise estimate, Peach Smoother uses its default, which is meant for use with a noisy TV capture. On any clean source material, that’s going to cause lots and lots of blurring.
If any part of the scene were standing still, Peach would have a chance to get a decent noise estimate, and the problem would be avoided. So there are a couple ways to take care of this problem, both of which are covered in the docs. I’ll reprint the FAQ on this, since it’s pretty relevant:
Quote:
I’m seeing lots of blurring in early parts of my constant motion (or very dark) video. What should I do to improve the results?
If you’re seeing lots of blurring, try using the Dot option. If the green dot doesn’t show up, then the problem is that the Peach wasn’t able to figure out how much noise there was.
There are a couple ways to solve this. The best is to turn on the Readout option, and watch a later part of the video where the picture is still. Make a note of the NoiseLevel and Baseline from that stationary part. Then specify them in your command, and the whole sequence should look fine.
If your video doesn’t have any stationary parts, then you should just skip this filter. A temporal smoother isn’t going to do much good for pure high motion material.
Another way to solve this problem is to put some still video (from the same clip) at the beginning of the sequence. That will allow the Peach to estimate the noise before the fast stuff shows up.
|
Note that fiddling with Stability and NoiseReduction isn’t going to help for this — The problem lies with the noise estimation, not the noise/motion tradeoff.
