View Full Version : unsmoothing?

jarthel
13th August 2004, 17:07
is there anything else other than unsmooth?

Thanks
Soulhunter
13th August 2004, 18:19
Originally posted by jarthel
is there anything else other than unsmooth?
Sure !!!

Beer, coconuts, carpets, lasers, table tennis, cars, whores and microsoft... :D

Seriously, what do you mean ???


Bye
jarthel
13th August 2004, 19:16
there are spatial/temporal smoothers. I have a video that has been overly smoothed.

I just want to introduce some sort of noise to unsmooth it.

I thought the mentioned of mf's unsmooth would be clear.

You even tested his unsmooth.
Soulhunter
13th August 2004, 19:29
Originally posted by jarthel
I thought the mentioned of mf's unsmooth would be clear.
My way to get some more info... ;)

Originally posted by jarthel
I just want to introduce some sort of noise to unsmooth it.
AddGrain introduces noise !!!


Bye
Mug Funky
14th August 2004, 18:30
best to do this on playback IMO. that is if it's possible

image processing is like whittling soap - everything you do to only serves to decrease the information in the picture, never increase it. adding noise is a pleasing effect (us humans are WEIRD), but only serves to add entropy - not real information. when encoding, too many bits will be used for the entropy, and there will be less for the actual image (what's left of it after all that smoothing).

however, if this is purely for aesthetic effect, and bitrate isn't a problem, then by all means add some noise :)

(it would be really good if some form of temporal deconvolution were available. this would be extremely useful for blended-field PAL transfers, in addition to restoring badly oversmoothed clips)
scharfis_brain
14th August 2004, 18:59
currently 24p and 30p in fieldblended PAL are (most of the time) recoverable, but 60i in PAL is NOT.

it is mathematically impossible...
Mug Funky
14th August 2004, 19:39
i understand that... sampling theory.

however, consider it like sample-rate conversion in audio:

48k to 44.1k can not be recovered completely, but can 44.1k to 48k? depending on how good the filters are, it should be. wouldn't it be the same for pixels in the same spatial location - 29.97hz from 50hz...

(am i barking completely up the wrong tree here?)