Ok so what am I looking at here.

[Simon Blackledge]

First still of 4k stuff on Red.com
Stills.. the MS of the plane..

Can I first say I just want to be clear what I'm seeing.. not bashing anything.. it's great.. just need to understand whats happening.

The grass.. the planes body.. basically at 100% it looks like a shot if I over do it in NeatImage or some noise reduction plugin. so is this

A, redcode compression
B, Jpeg compression
C, A+B = eek,,
D, Something Weta did.
E, something your working on.

Just an observation..

well done... ;)

si

[GlennChan]

Ok my question. :D

In the first jpeg (002318.jpg), what's the weird purple/green stuff on the edges of the plane with the sky, on the wires connecting the wings, on the text on the plane, etc. ?

[Simon Blackledge]

possibly CA ?
http://en.wikipedia.org/wiki/Chromatic_aberration



are current lenses good enough ? mtf etc? to keep up with 4k sensors?

[Floris Liesker]

It's very likely a noise reduction done in the Pablo grading.

See also http://www.reduser.net/forum/showthread.php?t=1944

[GlennChan]

The direction of the green/purple fringing doesn't look consistent with CA. i.e. there are circles (i.e. the crosshairs on the plane) and u-shaped areas in which the inside of that is all green/magenta.

[Jonathan Payne]

The magenta fringe on one side and green fringe on the other are indeed characteristics of Chromatic aberration. This is most likely what we're seeing in the image. That said, there is some degree of CA in all lenses and because of how film works it helps to hide CA while it becomes very noticeable in all digital sensors. The finest still cameras all show varying degrees of CA. No lenses can truly eliminate it 100% but their are several ways to combat it.

And this should not be viewed as problem with the camera. It's likely due to the lens and sensor combination. Not all combinations work together well.

[Mike Prevette]

TO explain that more it has to do with the wavelengths of light all focusing on a slightly different plane. Film corrects for this by making it's chromatic sensitivity in layers. Digital sensors all lie on the same plane hence more pronounced CA.

_mike

[Miguel "Macgregor" De Olaso]

First still of 4k stuff on Red.com
Stills.. the MS of the plane..

Can I first say I just want to be clear what I'm seeing.. not bashing anything.. it's great.. just need to understand whats happening.

The grass.. the planes body.. basically at 100% it looks like a shot if I over do it in NeatImage or some noise reduction plugin. so is this

A, redcode compression
B, Jpeg compression
C, A+B = eek,,
D, Something Weta did.
E, something your working on.

Just an observation..

well done... ;)

si

I am worried about this too. It is the same "problem" i see with Dalsa cameras. And that is the reason why i dont like Dalsa cameras.

Noise reduction? Demosaic? It does not look like jpeg compression.
It looks like the 1st Kodak digital photo cameras that claimed 5Mp having sensors of 3 or 4 Mp. They had a lower native resolution and filled the empty spaces.
However the Mysterium sensor should have more than enought pixels to play with it.
An answer is needed, but certainly i dont like it. DSLRs dont look like this. Film does not look like this.

[Clayton Harper]

Ok my question. :D

In the first jpeg (002318.jpg), what's the weird purple/green stuff on the edges of the plane with the sky, on the wires connecting the wings, on the text on the plane, etc. ?

It sort of looks like someone pulled a really bad key on a DSLR file. :bye2:

[Miltos Pilalitos]

I think you guys should really see how similar a 2K scanned film frame can look when you inspect it as much as you have.

Do you think the CA on the plane's edge with the sky and around the text is BAD? Wait until you see CA from panavision film cameras with anamorphic lenses or to be more fair from Arri 535 with Zeiss lenses.

99% of your audience will never notice this in such collosal resolutions!

The waterpaint effect could be a combination of things. Prototype camera, post processed footage, jpeg compression.

i think that you worry too much for nothing.

[Stokestack]

Possible CA is not what flameop and other people are worried about. The concern is artifacts. I posted some still-frame crops for comparison in this thread:

http://www.reduser.net/forum/showthread.php?t=1944&page=14

[Clayton Harper]

To be fair, I imagine some tweaks to the debayer code could probably resolve these artifact issues. Let's have a little faith in the power of Nattress.

[Alexander Black]

They need to post TIFFs - and they need to put out some raw frames as shot. That would give people a much, much better handle on what was coming out of the camera. :)

_a

[Brook Willard]

I think you guys should really see how similar a 2K scanned film frame can look when you inspect it as much as you have.


Sounds like a good idea. Enjoy. (http://homepage.mac.com/brookwillard/2K5218old.jpg).

[Simon Blackledge]

When I was using a very early version of Magic Bullet, and it was de-interlacing digibeta, and came across grass or a tree this is the pattern effect I would see. Never really saw it on playback though after.

It looks like something just needs tweaking because latest MagicBullet is way better at this now.. it will improve.

Also look how much more detailed film looks on the fly due to grain than it does on a still.

Apologies if I started anything..
s

[Mike Zinner]

I guess the reason why I was so shocked seeing the PJ stills is, that up until now all the pre-alpha footage looked absolute perfect to me.

* The BubbleGum shot and still (still @ 4K)
* The MilkGirl Shots and still (4K)
* The Porsche and Watch shot (ok, only 1K)
* The GreenScreen test stills (4K)

They just looked like a series of stills from my Nikon D50, with even higher resolution. So RED has already proven that then sensor is capable of shooting these absolutely gorgeous images and clips.

So I start wondering what has been changed in respect to the Pre-Alpha stuff so that the stills do not look nearly as good as they already did (I mean, let me repeat that, Pre-Alpha stuff already looked perfect to me).

And I guess that is a lot that has been changed. But the only things I know for sure are:
A) now REDCODE RAW was used and that the Pre-Alpha footage was recorded directly from the RED RAW PORT
B) there has been some heavy post production performed on the footage

I don't think A) can be the reason, since it's a visually lossless codec. That is also proven by the fact that the only other RAW codec I know, Cineform RAW, isn't suffering from any of the issues we see in the new Alpha stills.

So, I think it's B). And this is something that can be easily addressed and it doesn't leave me worried at all.

I am stopping any speculations/muse about it on my side for now, forget about the Alpha images and just give the RED team more time to sort things out and wait for Beta and GA footage being posted. As simple as that.

Mike

p.s.: I was much more worried about the motion rendering. But as the "Crossing the Line" clip has shown there is absolutely nothing to worry about, it's gorgeous.

[Simon Dean]

I posted over on the other thread about this.

Really, to answer your question, I don't think anyone can no what the problem truly is as the workflow was not under completely controlled conditions...and that's what I think it is down to - the workflow.

The thing to learn is that with footage that starts so perfect from Red, everything you do to it gives the possibility to make it worse. This is really well known in audio recording studios - perfect mic, perfect pre-amp, dodgy cable = dodgy sound. (And there are many other steps that could mess it up).

Once production cameras are out, people will be trying all kinds of workflow and then we'll know which the best ones are.

[Nick Shaw]

I think it's important to remember that they were 'only' monitoring at 2k for the grade. So settings that they used which gave a look that they liked at 2k, might not look so nice when examined closely pixel for pixel at 4k. If you take the bi-plane image and reduce it to 50% then view that at 1:1 you can see how artefacts introduced by the grade could have gone un-noticed.

[Simon Blackledge]

Agree, it looks to me like a post-process of sorts has been applied rather than a compression algorithm. But hey ho.. just dropped the clip on the appletv and viewed it on a 42" plasma.. ace ;)

[Floris Liesker]

I think it's important to remember that they were 'only' monitoring at 2k for the grade. So settings that they used which gave a look that they liked at 2k, might not look so nice when examined closely pixel for pixel at 4k. If you take the bi-plane image and reduce it to 50% then view that at 1:1 you can see how artefacts introduced by the grade could have gone un-noticed.

That must indeed be it, Nick. Very logical explanation. At 50 % the artifacts are as good as gone, so I think they used a little too agressive denoising algorithm without being able to see the effects of it at 4K. Good thinking.

[Simon Blackledge]

..but it was shown at 4K?

How big was the screen @ NAB ?

[Simon Dean]

Good point - I'm still thinking that something happened to it after that. What format was it played in (I.e. what player and what codec) and what was the method for extracting the stills? Do we know what they used to play it?

[Álex Montoya]

I am absolutely confident that the artifacts' issue will be taken care of in time, but right now I am really interested in Graeme's opinion an what may be going on with these pics, since he's the expert here and has already stated that he hasn't seen anything like this in previous footage.

[Brook Willard]

..but it was shown at 4K?

How big was the screen @ NAB ?
Yes, 16'x9'

[Miguel "Macgregor" De Olaso]

Why this smoothness does not appear on the first images and why does it appear here?
This is the most important question of all.

[Simon Blackledge]

Yes, 16'x9'

thanks Brook.

Well the 1st images never went through Quantel maybe ... ;-)

[Gavin Greenwalt]

I think it's important to remember that they were 'only' monitoring at 2k for the grade. So settings that they used which gave a look that they liked at 2k, might not look so nice when examined closely pixel for pixel at 4k. If you take the bi-plane image and reduce it to 50% then view that at 1:1 you can see how artefacts introduced by the grade could have gone un-noticed.

This actually tells me a great deal. If it was a noise reduction filter you always have to be careful with the 1:1. It's often really easy to accidentally go too far when you're zoomed out and you don't realize what you're losing.

Like I said in another post there seems to be a really quite good 2k image in those stills. It's the 4k that falls flat.