Adding more filters (degrainmedian) before seesaw

[Moti172]

Hello

Is there a sense in adding the following filters before seesaw, in order to reduce the grain in a movie?

degrainmedian(mode=2)
degrainmedian(mode=2)
a = last
b=a.degrainmedian(mode=1)
SeeSaw(a,b, NRlimit=5, NRlimit2=6, Sstr=1.5, Slimit=5, Spower=5, Sdamplo=6, Szp=16)

I tried to use only seesaw for improving the sharpness and to remove grain, but it wasn't enough.

Thank

[Didée]

You can do so, sure. But the more denoising is done before, the more the SeeSaw principle will be canceled: If the "original" (as delivered to SeeSaw) has been denoised already, then any detail lost at this point can't be recovered by SeeSaw.


Usually, one should try to use the unaltered original for "a", and an sufficiently strong denoised clip for "b". Then, various settings of SeeSaw can be adjusted until the result is calm enough:

make NRlimit is big enough (too low numbers will apply not enough of the denoising achieved by "b"), lower "bias", raise "SdampLo", use "sootheS" (spatial soothing) in combination with bigger and/or negative values for "sootheT" (negative numbers will do double temporal soothing)

Compare: in this case (http://forum.doom9.org/showthread.php?p=751643#post751643) of a very noisy clip I used the following settings:

a = last
b = mcnr_simple2(frames=3,thY=24,thC=20,thY2=-1,thC2=-1,removdirt=true,LPrad=2.0,LPlosens=1,LPhisens=5)

SeeSaw(a,b, NRlimit=15,bias=32,Smode=11,Szp=25,SdampLo=20,sootheT=-49,sootheS=86) # red: syntax of newer SeeSaw

The "a" clip was kept with all of its noise, and the "b" clip had been denoised *very* strong.
SeeSaw then got
- a big NRlimit, because so much "maximum pixel change" had to be allowed in order to remove the noise.
- a lowered bias, to not use too much noise from the original for sharpening
- a rather big SdampLo, because with such a noisy source it's better to not amplify very small sharpening effects
- a big sootheS to reduce "hot pixels" coming out of the sharpening
- negative sootheT to make sharpening result temporally more stable
- Szp was raised a little, to compensate a little for all the other "reduced" settings


Playing with the parameters and with the external denoising often can make very big differences.

If all of that isn't enough, however, one can additionally use some moderate denoising prior to the a/b/SeeSaw part ...
... but on that road, one may come very quickly to a point where it simply makes no sense anymore to use SeeSaw, and the old game of source.Denoise().Sharpen() would deal just as well. :)

[Moti172]

Thank you for the answer

I'll read the seesaw instruction again and try to change the seesaw's parameters.