[Tig] Noise Reduction: where in pipeline?
patrick at bluegiraffe.tv
Fri Jun 10 06:17:01 BST 2011
Craig Level wrote:
I am now damn sure that we will hear from FilmMaster supporters.
I thought you'd be disappointed if you didn't ....
Sent from my android device.
From: Craig Leffel <craig at optimus.com>
To: Richard Kirk <richard at filmlight.ltd.uk>
Cc: "tig at colorist.org" <tig at colorist.org>
Sent: Fri, 10 Jun 2011 5:38
Subject: Re: [Tig] Noise Reduction: where in pipeline?
Sohonet http://www.sohonet.co.uk sponsors the TIG.
Support from Nucoda www.imagesystems.tv
The TIG thanks Michael Wilsker for support.
Hello Richard - and all the Tig peoples -
As one of the people that uses that little Temp Degrainer you guys built....
All I can say is that there is way too much emphasis on "grain" reduction, and not enough on "noise reduction". Grain is grain. It's either present in offending quantities or it's not. Sure, it has different structures different inherent properties... but it's GRAIN.
What commonly gets forgotten in a commercially built, widely used, generic-ish, good for "everyone" color corrector -
is that production "happens" on a really wide variety of cameras... on a variety of media, and on a variety of intended devices.
I've been having this discussion since I was color correcting my first spots where 95% of the spot was 35mm Film, and a few shots were from VHS.... for SD. Now that's been replaced by Film mixed with Quicktime, 7D, Phantom, Red..... etc. for Broadcast or Web.... at least in my world, I'm not a colorist for feature films.
The point is we NEED a noise reducer these days even more than a grain reducer. I encounter noise that builds from the moment the image is captured to the moment I output it to its intended record media or render media. I often do a separate NR for a rendered strip to save time and get better NR. ( I should say TG ).
One of my biggest problems is 7D. Alexa with LUTs is a problem too. So is Red. Omg and then theres Phantom... Holy sheepshit. All different media, ALL different issues.
I mean no disrespect to Baselight. It's an amazing product that I make great pictures on ( and a nice living too!)
This problem is inherent in the Film vs All Else argument and is true for almost any widely available Color Corrector.
I am now damn sure that we will hear from FilmMaster supporters, but I would argue the same point. Noise is not grain. Multiple kinds and qualities of Noise caused by capture, color correction and sharpening or plugins is not one kind of Noise, but many blended together into a complex problem that has no easy answer.
However, I'm sure people are slaving away on code right now to try and get a better handle on it.
Good luck to those programmers in the trenches.
On 6/9/11 3:05 AM, Richard Kirk wrote:
> Sohonet http://www.sohonet.co.uk sponsors the TIG.
> Support from Nucoda www.imagesystems.tv
> The TIG thanks Michael Wilsker for support.
> In theory you could make a degrainer work in any colour space. However, having written a temporal degrainer, I would always recommend people to degrain log images before you do anything else. If you are writing a degrainer, you usually set the default settings so it assumes a 'grain' is a roughly circular roughly pixel sized object with soft edges that does not appear on the adjacent frames, and a constant density contrast. There will be a threshold parameter, and probably some specialist settings for black and white images, very quiet images, anamorphic images, and so on. However, the closer you get to the 'just scanned' image, the more likely it is you will get the results that the degrainer writer intended.
> In practice, I see the temporal degrainer used at the exact opposite end of the pipeline. This is natural, I guess: you grade the image, bring out the highlights, comp in the features, and then think "you know what - the sky is a bit noisy....".
> Could be worse. I have had people using it to clean up interlaced video.
> Richard Kirk
More information about the Tig