HDRx Noise/Artifacts

[Mark Andersen]

Recently in I shot and evening cityscape at 9:1 (accidentally, was working too fast) although the shot is useable I could see some compression artifacting due to the fine detail in the shot. I wished I had used less compression in this case.

 

[Justin Kirchhoff]

I've had this issue in the exact same setting. At least for mine it's such a quick shot that it's almost unnoticeable. I'd go back and reshoot it. Why not?

 

[Filip Orlandic]

Hey,

It's normal to see those artifacts in dark-black areas due to a high compression ratio.

Has anybody ever shot at RC28 on RED ONE? Did You forget how it looks like? Try to remember...



TEST June, 2011.

BLUE CHANNEL


EPIC-M RC 4:1

RED ONE M-X RC 8:1

EPIC really works impressive at RC 3-5:1 and it's clean like an uncompressed footage. Trust me.

Sometimes there are situations where it falls of course but due to a high compression ratios (over RC 6:1).

BTW,
I've already archived all of my NEF/TIFF photos in jpeg2000 4:1 compression (7:1 was my max limit). Above 7:1 it started to eat my photos



EPIC 5120x2700 single uncompressed RAW frame should be around 26MB. You guys are shooting at 2-4MB file sizes. S*** must happen sometimes.

I think with a bit faster SSDs we might be able to shoot better compession in slow motion. Instead of shooting 96fps at RC 12 we could shoot at RC 10.

Based on my calculations...

 

[Michael Millichamp]

I've had this issue in the exact same setting. At least for mine it's such a quick shot that it's almost unnoticeable. I'd go back and reshoot it. Why not?

The main reason was because I still had no external power, but I got notification my sidehandle will be here tomorrow, so I guess I will go reshoot.

 

[Nook Kim]

Hey,

It's normal to see those artifacts in dark-black areas due to a high compression ratio.

Has anybody ever shot at RC28 on RED ONE? Did You forget how it looks like? Try to remember...



TEST June, 2011.

BLUE CHANNEL


EPIC-M RC 4:1

RED ONE M-X RC 8:1

EPIC really works impressive at RC 3-5:1 and it's clean like an uncompressed footage. Trust me.

Sometimes there are situations where it falls of course but due to a high compression ratios (over RC 6:1).

BTW,
I've already archived all of my NEF/TIFF photos in jpeg2000 4:1 compression (7:1 was my max limit). Above 7:1 it started to eat my photos



EPIC 5120x2700 single uncompressed RAW frame should be around 26MB. You guys are shooting at 2-4MB file sizes. S*** must happen sometimes.

I think with a bit faster SSDs we might be able to shoot better compession in slow motion. Instead of shooting 96fps at RC 12 we could shoot at RC 10.

Based on my calculations...

Yeah.. What you say almost makes sense, but it really doesn't make sense when a 5D outperforms in compression artifacts when 5D compresses the hell out of the image.

 

[Mike Lary]

Yeah.. What you say almost makes sense, but it really doesn't make sense when a 5D outperforms in compression artifacts when 5D compresses the hell out of the image.

Here's a specific example of what Nook is talking about. This was from a multi camera shoot (Epic, Phantom, 5D, 7D). Epic's compression was by far the worst of the lot. You can see a snowstorm-like pattern of artifacts through the entire image, most noticeable in darker tones.
https://www.yousendit.com/download/M3BuMWZOUnFFc0xyZHNUQw
This is a cropped portion (we can't show faces) of a properly exposed R3D (2K, 240fps, 8:1) exported from RCXpro as ProRes 4444. If I can get RCXPro to properly export a cropped R3D I'll post one here.

I'll be happy to post samples from all cameras if anyone is interested, but the real issue is that the footage is completely unacceptable by professional standards regardless of the camera used or the processing effort required to produce images at a given frame rate, resolution, etc.

 

[Matt Ryan]

Here's a specific example of what Nook is talking about. This was from a multi camera shoot (Epic, Phantom, 5D, 7D). Epic's compression was by far the worst of the lot. You can see a snowstorm-like pattern of artifacts through the entire image, most noticeable in darker tones.
https://www.yousendit.com/download/M3BuMWZOUnFFc0xyZHNUQw
This is a cropped portion (we can't show faces) of a properly exposed R3D (2K, 240fps, 8:1) exported from RCXpro as ProRes 4444. If I can get RCXPro to properly export a cropped R3D I'll post one here.

I'll be happy to post samples from all cameras if anyone is interested, but the real issue is that the footage is completely unacceptable by professional standards regardless of the camera used or the processing effort required to produce images at a given frame rate, resolution, etc.

Dumb question but did you black shade?

I checked out the clip and it looks terrible. I've never seen that before.

 

[Mike Lary]

Dumb question but did you black shade?

I checked out the clip and it looks terrible. I've never seen that before.

Yes, we did a successful black shade calibration. We also shot footage the next day and had no problems. One disturbing thing about this kind problem is that it's not consistent. Sometimes it comes and goes within one shoot, spanning several clips on multiple SSDs. In this case it happened continuously throughout the entire shoot (20 hours in total).

I'm glad you say the clip looks terrible. Red support told me it was very hard to see exactly what was wrong without seeing a side-by-side comparison with "good" 2K footage shot under the same conditions.

 

[Matt Ryan]

Yes, we did a successful black shade calibration. We also shot footage the next day and had no problems. One disturbing thing about this kind problem is that it's not consistent. Sometimes it comes and goes within one shoot, spanning several clips on multiple SSDs. In this case it happened continuously throughout the entire shoot (20 hours in total).

I'm glad you say the clip looks terrible. Red support told me it was very hard to see exactly what was wrong without seeing a side-by-side comparison with "good" 2K footage shot under the same conditions.

I've shot 2k footage in low light with this much noise. It doesn't look right at all. Hope this gets figured out. Can you take a grab of your histogram? Also what are your meta settings?

 

[Mike Lary]

Can you take a grab of your histogram? Also what are your meta settings?

Sure. Here's a histogram from the full frame.

By digitalfactory at 2012-03-16[/IMG]

Not sure what you're looking for in metadata, but here are some details:


2048x1080
23.976 project timebase
240 fps
REDcode 8:1
No HDR
EPIC-X
ISO 800
REDgamma2
REDcolor2
1/480 shutter
Shutter(ms) 2
5600K

 

[Matt Ryan]

Everything is where it should be settings and exposure wise. Thats really weird. You say this comes and goes at random?

Have you by chance tried reseating the ssd module?

 

[JanneJansson]

Yeah.. What you say almost makes sense, but it really doesn't make sense when a 5D outperforms in compression artifacts when 5D compresses the hell out of the image.

What you say makes almost no sens. 5D have many, many quality issues not coparable to Epic at all. Shooting digital is not shooting film where you have every silver molecules attention on the film surface. All digital cameras have a dynamic range and also a data bandwith. You have to pay close attention on both so make a great image. Some factors that affect the end quality is always.

- Compression ratio (If you chose to throw away data then you image will fall apart more)
- Noise/ISO (less noise to encode, the more bandwith to image detail) In some cases noise use up very much data.
- Detail in shot. (shooting a very detailed image require more data)
- FPS (more frames per sek equals less data to every frame)

When shooting at higher framerates the sensor also starts to deviate from baseline and usually require more light just to compensate for this. Shooting hi detailed image with hi compression will make you image have compression atifacts. Shooting hi-detailed scen in hispeed with hi compression in lowlight will make your 5K look like VHS.

 

[Nook Kim]

What you say makes almost no sens. 5D have many, many quality issues not coparable to Epic at all. Shooting digital is not shooting film where you have every silver molecules attention on the film surface. All digital cameras have a dynamic range and also a data bandwith. You have to pay close attention on both so make a great image. Some factors that affect the end quality is always.

- Compression ratio (If you chose to throw away data then you image will fall apart more)
- Noise/ISO (less noise to encode, the more bandwith to image detail) In some cases noise use up very much data.
- Detail in shot. (shooting a very detailed image require more data)
- FPS (more frames per sek equals less data to every frame)

When shooting at higher framerates the sensor also starts to deviate from baseline and usually require more light just to compensate for this. Shooting hi detailed image with hi compression will make you image have compression atifacts. Shooting hi-detailed scen in hispeed with hi compression in lowlight will make your 5K look like VHS.

Hi JJ,

You need to take a look at the link provided by Mike Lary above, and please tell me if any of your points are applicable to the case. It wasn't high detailed scene, the 8:1 compression ratio shouldn't be too much of compression setting for the given scene, the scene was properly exposed with enough light, and all other cameras performed much much better than the Epic. Please ask yourself if you would accept this quality from your camera.. I mean any video camera. I have never seen this kind of artifacts at the acquisition stage on any digital cameras.

I agree Epic is supposed to be a much better camera than 5D, but unfortunately it wasn't the case in this case.

 

[Mike Lary]

Everything is where it should be settings and exposure wise. Thats really weird. You say this comes and goes at random?

Have you by chance tried reseating the ssd module?

There was a noise thread in which a couple users reported the same issue. Sometimes one bad clip is sandwiched between good clips, but other times the bad clips span multiple SSDs. In this case 20+ hours of footage was shot, so the SSDs were reseated several times. On-set troubleshooting isn't likely to happen with this kind of issue because the operator isn't likely to notice the artifacts on the LCD.

When we first saw this issue, we were shooting varispeed at all resolutions. The issue was harder to see on 5K, but it was still there if the conditions were right (out of focus darker tones - tree trunks, etc). It's possible that the issue doesn't so much come and go as it's harder to see under different lighting conditions.