PDA

View Full Version : SmoothUV bug report


Kurosu
2nd September 2003, 20:37
Originally posted by Dreassica
Id be gratefull if u "unbroke" radi higher as 4, cuz i still can use those with some clever tweaking of the settings, since i do use it for extremely noisy anime sources. I didnt see any Maintain diffweight- ish parameter in the filter as well. Any reason for that?
Its definitely faster for me as smothhiq was, now a yv12 tempsmoother and im totally happy..

BTW it does work fine in qmf, i just jumped to conclusions too fast. :)

1) Are the radius (radiuses? radii?) higher than 4 really broken? I've tested 5 and 6 which use a different processing, and I couldn't see any problem where I would have expected them.
If you are talking of the fact radius higher than 7 can't be selected, then this is intended. I don't plan to had radius 7 (that's already a 15x15 kernel) because it's too much hassle for the benefit over radius 6, and it would only be used for that very case.

2) I wrote on my page that it was a simplistic remake of smartsmoother. I've added a part which makes it of higher quality but not HiQ :)
if you properly read the page once again, you'll see the booleans HQY and HQC activate the hiqh quality mode for respectively Y channel and U,V channels. To go deeper into details, it almost works like msmooth: it calculates a local edge value for the current pixel and plane processed, and decrease accordingly the weigh of the smartsmoothed pixel. That leads us to the parameter "strength": the higher it is, the more you'll keep the smartsmoothed version and keep the original. This strength is dynamically modified by edge value in HQ mode: the higher the edge, the less the smartsmoother pixel will be used. The current setting makes it worth for anime I bet, as it is very sensitive to edge:
new strength = input strength - 4*edge value
Check by yourself how the HQ mode changes the output (strength isn't used in none-HQ mode).

3) tempsmoother? Like the one in the old mpeg2dec? temporalsoften isn't sufficient ?

Dreassica
2nd September 2003, 21:07
from the readme that came with the filter: Radii > 4 (ie 5 and up) are busted. I leave them available for you to check
but consider that a radius of 4 produces a 9x9 convolution kernel. It should
be pretty sufficient for most uses.

Thtas why i was asking it, but if u say it doesnt pose any problems to use them, i'll beleive ur word for it. as for radii 7+ i wasnt really talking about those, more likr 5,6. Any higher would be useless.

I did mean ol tempsmother from mpeg2dec.dll, i tried Tsoften on it, result werent really satisfactory, didnt clean up the noise well. Tempsmoother still does a better job i think on it.

Kurosu
2nd September 2003, 22:16
OK, bug found. Located in "SSmooth.txt" file :)
Updated package together with a 1.4 version, where radius 7 is added, and HQ mode for radius > 4 really does something (wasn't previously).

Dreassica
2nd September 2003, 22:59
I f i set luma radius 4 or higher i get artifacts in some frames, which arent there if i set itt lower than 4.
http://www.geocities.com/dreassic/uvsmoothartifact.JPG

ps u ned to paste the link in a browser, clicking wont work.

Kurosu
2nd September 2003, 23:42
I *may* have found the bug, but that means there are many filters where it is, so I'm updating them at the same time.