Heavy Compression Artefacts

[Rainer Fritz]

Hi !

Had a testshoot yesterday with an EPIC-M and Masterprimes. One shot shows strong compression artefacts.
Interesting thing is, that it is stronger when the camera and actor is not moving (black cloth).
The picture has a lot of detail in it (woods). Look to the beginning and end of the clip. The R3D file can be downloaded via ftp:

URL: ftp://dcs.co.at
User: web280f5
PW: 123456

hope for feedback...
best
rainer

 

[Rainer Fritz]

bump

 

[Robert Berger]

Hi Rainer, saw the shot with the girl, maybe a good idea to send it to RED for a closer look. you shot 6:1 compression and that should be oke!
Yous slomo shot on 2K must be done at as lowest ISO the 250 iso othrewise you will have a lot of noise in your shot.

Best regards
Robert M. Berger NSC
Epic-X "ED" 00274 and R1 "FRED" 1265 MX SSD 1265
The Netherlands

 

[Tom.Wong]

I see what you're talking about but it's hard to tell whether it's just a combination of how the shadows are falling and the wrinkles in the material. the shadows on the branches are fine, as does the darkest material on the scarf. are you seeing this in any other shots? is this repeatable but with a large solid piece of black?

 

[Rainer Fritz]

I filed a support ticket...

Tom: In the black clothing of her you can see the artefacts clearly, but the rest of the picture shows it also before and after the pan. I saw this only that strong in this particular shot. But we need to know if this could happen again.

rainer

 

[JanneJansson]

This is a hard scene for the codec to handle. You need to help the camera here.
- Minimize the ratio, all data you can give.
- More light, (rate camera at 500iso) Usually not good, but better then compression patterns.
- If you still have compression patterns, you need to lower the resolution untill you have a clean frame.

 

[Rainer Fritz]

because the sensor has that size I'm thinking of shooting in 4k 6:1 and as you said exterior with 500ISO....

 

[Tom.Wong]

iso really means nothing, you are shooting raw. the sensor is always recording at 320, and dialing in a higher iso is just gain at that point. you were shooting 6:1, 5:1 might help that much more, but in the case of this 3:1 might be necessary. if you are doing any overcranking than only suggestion is to pump more light in the scene if Red determines nothing is wrong with the camera itself.

 

[Adrian T.]

The closer the recorded image comes to random noise the more lossy the compression will be. That's the problem with any lossy compression.

I made a test shooting a plate consisting of a random noise pattern with a little nice gradient and a black "X" placed in its center. The "X" and the gradient showed extreme artifacts at "normal" compression rates but still at 4:1 they were far from being clean. And when the clip was running you cloud see that the whole picture was extremely noisy too: another hint that compression was lossy (if your exposure was ok of course).

In your example all the little branches make up a pretty "random noise" when the camera is still. As soon as you pan, the motion blur on the background reduces all the visual details and compression quality increases.

It would be great if Graeme could chime in here. Maybe there is some more tweaking of compression parameters that could be done on RED's side... Apart from that: always try shoot with the best compression rate available in situations like these.

 

[Jannard]

I haven't downloaded your footage but so many projects have been shot at 6:1 with no issues I'd suspect that something else is going on. RED ONE MX (like The Girl with the Dragon Tattoo) is shot at the equivalent of 7.5:1.

Almost every EPIC project (Spiderman, The Hobbit, Prometheus, etc) has been shot at 5:1 or 6:1 with no issues. This includes stuff for VFX. The bomb squad needs to get involved if you think there is a problem.

Jim

 

[Rainer Fritz]

Thx Jim for responding. Yes I'm in contact with Brian and uploaded the first half of the shot on rufus.
I know that the cam is performing well on a big number of productions and thats good.
As said it was only in this particular shot and it will be interesting I think for your technicians as well to look into it.
probably there is something which can be tuned, but when you look to the clip you will see there is something going on.

rainer

 

[Adrian T.]

This is a 1:1 crop of my compression test at 4:1 I mentioned in my post above:

Of course it's rare that you shoot an image like this in the real world. It simply shows the limitations of lossy compression with a fixed bandwidth: increased noise and artifacts, decreased sharpness. It's a mathematical fact. There's not much that can be done about it. We all have to live with it.

 

[Will Keir]

Adrian,

Was there a final verdict on what might have caused the issue? I'm working though similar issues myself.

"Live with it?"

Best.