Topics

Camera noise measurements

Robert Payton
 

Love your clarity of thought Geoff - i was wondering why people  didnt get it !

Rob

DoP Director  Cape Town

On 6 Feb 2018, at 08:23, Geoff Boyle <geoff.cml@...> wrote:

Lower is better in terms of the numbers in this test.
Is lower better visually, well that’s a personal thing.
The attempt here is to give comparative numbers, i.e. 10 is lower than 20.
It’s simple, there is more noise in an image that requires a setting of 20 in NR than there is in an image that requires 10.
It’s a simple test, at what point are the lines on the WF clear?
It’s a test that anyone can try without access to expensive test kit.
Download the test files in EXR format and then see what level of noise reduction YOU like when you pull up a 4 stops underexposed picture.
 
Simples!
 
Cheers
 
Geoff
DP
Netherlands
 
From: cml-general@... [mailto:cml-general@...] On Behalf Of Tim Sassoon via Cml.News
Sent: Tuesday, 6 February 2018 02:21
To: cml-general@...
Cc: Timothy Sassoon <tsassoon@...>
Subject: Re: [general] Camera noise measurements
 
IS lower better in all cases? Or is the judicious use of DNR a reasonable strategy in search of other desirable properties?
 
 
Tim Sassoon
Venice, CA
 
 
 
 
On Feb 5, 2018, at 9:22 AM, Geoff Boyle <geoff.cml@...> wrote:
 
Lower is better.
 

Nick Morrison
 

Helium was shot 7K right? To keep it Super 35?

If so, worth mentioning it would presumably be less noisy at 8K...esp once scaled down to UHD...

___________
Nick Morrison
Founder, Director & Lead Creative
nick@...
646-236-7884
smallgiant.tv


Paul Curtis
 

On 6 Feb 2018, at 08:04, Argyris_Theos_cml <cml@...> wrote:
My question (not objection, just question) is: is any software powered waveform good enough for noise evaluation?
Or perhaps it would be better to do it in a hardware waveform monitor?
I was told once that Resolve adds a small dither to the images and therefore would make testing for this very difficult (circa v9 anyway). However the caveat is i cannot verify that as fact (i don't use Resolve to test anyway)

Also in movement different styles of noise are aesthetically different, so it's pretty tough to reduce that to a number.

There are other empirical approaches:

http://www.imatest.com/solutions/iso-15739/

That seems to accommodate different tones and is super comprehensive, but to what real world useful point i'm not sure.

As Geoff says we can download ourselves and see...

cheers
Paul

Paul Curtis, VFX & Post | Canterbury, UK

Colin Elves
 

Are we reinventing the wheel a bit here? Can’t RMS Noise and SNR be used to put an objective number on this? 

Also, surely the issue is more dynamic range than simply ‘noise’. What’s important is the difference between peak signal and RMS Noise? Right? 

Or have I misunderstood all this?

Colin Elves
Director of Photography
Berlin/London




On 6 Feb 2018, at 09:04, Argyris_Theos_cml <cml@...> wrote:

My question (not objection, just question) is: is any software powered waveform good enough for noise evaluation?
Or perhaps it would be better to do it in a hardware waveform monitor?

Argyris Theos, gsc
DoP, Athens Greece,
theos@...
+306944725315
Skype Argyris.Theos
www.vimeo.com/argyristheos
via iPhone

6 Φεβ 2018, 8:23 π.μ., ο/η "Geoff Boyle" <geoff.cml@...> έγραψε:

without access to expensive test kit.

Argyris_Theos_cml
 

My question (not objection, just question) is: is any software powered waveform good enough for noise evaluation?
Or perhaps it would be better to do it in a hardware waveform monitor?

Argyris Theos, gsc
DoP, Athens Greece,
theos@...
+306944725315
Skype Argyris.Theos
www.vimeo.com/argyristheos
via iPhone

6 Φεβ 2018, 8:23 π.μ., ο/η "Geoff Boyle" <geoff.cml@...> έγραψε:

without access to expensive test kit.

Geoff Boyle
 

Lower is better in terms of the numbers in this test.

Is lower better visually, well that’s a personal thing.

The attempt here is to give comparative numbers, i.e. 10 is lower than 20.

It’s simple, there is more noise in an image that requires a setting of 20 in NR than there is in an image that requires 10.

It’s a simple test, at what point are the lines on the WF clear?

It’s a test that anyone can try without access to expensive test kit.

Download the test files in EXR format and then see what level of noise reduction YOU like when you pull up a 4 stops underexposed picture.

 

Simples!

 

Cheers

 

Geoff

DP

Netherlands

 

From: cml-general@... [mailto:cml-general@...] On Behalf Of Tim Sassoon via Cml.News
Sent: Tuesday, 6 February 2018 02:21
To: cml-general@...
Cc: Timothy Sassoon <tsassoon@...>
Subject: Re: [general] Camera noise measurements

 

IS lower better in all cases? Or is the judicious use of DNR a reasonable strategy in search of other desirable properties?

 

 

Tim Sassoon
Venice, CA

 

 

 

 

On Feb 5, 2018, at 9:22 AM, Geoff Boyle <geoff.cml@...> wrote:

 

Lower is better.

Tim Sassoon
 

IS lower better in all cases? Or is the judicious use of DNR a reasonable strategy in search of other desirable properties?


Tim Sassoon
Venice, CA




On Feb 5, 2018, at 9:22 AM, Geoff Boyle <geoff.cml@...> wrote:

Lower is better.

Jeff Kreines
 

Gee, I remember when camera noise was measured 3 feet from the lens, and 30 dB was noisy, 20 dB was wonderful.


Jeff Kreines
Kinetta
jeff@...
kinetta.com

Geoff Boyle
 

All scaled to UHD

cheers
Geoff Boyle


From: cml-general@... <cml-general@...> on behalf of Gavin Greenwalt <im.thatoneguy@...>
Sent: Monday, February 5, 2018 6:52:01 PM
To: cml-general@...
Subject: Re: [general] Camera noise measurements
 

Did you rescale all the images to the same resolution? Depending on how the waveform operates it might not be averaging pixels which means it’s missing what the eye does naturally looking at a high resolution display or what math does when you downsample.

Gavin Greenwalt
VFX Sueprvisor
Seattle, WA

 

Gavin Greenwalt
 

Did you rescale all the images to the same resolution? Depending on how the waveform operates it might not be averaging pixels which means it’s missing what the eye does naturally looking at a high resolution display or what math does when you downsample.

Gavin Greenwalt
VFX Sueprvisor
Seattle, WA

 

deanan@gmail.com
 

Very nice way to test for noise!

Quick note: if a camera is being offset/lifted less than another, then it's starting out at a higher level of exposure which means it'll have less noise from the exposure. Do you happen to have any lens cap shots of total black?

That'll potentially give you misleading results. 
Not saying the results aren't practically viable, but it does add some variability to the facts...


Deanan DaSilva
Consultant
Playa del Rey, CA

On Mon, Feb 5, 2018 at 9:22 AM, Geoff Boyle <geoff.cml@...> wrote:
Lower is better.

cheers
Geoff Boyle


From: cml-general@... <cml-general@...> on behalf of Adrian Wolfson <adrian@...>
Sent: Monday, February 5, 2018 6:03:21 PM
To: cml-general@...
Subject: Re: [general] Camera noise measurements
 
Hi Geoff
I would really like to understand what your test actually means.
Are you saying Varicam has lowest noise level or Alexa?
Is green gain lower better?
Is NR values lower or higher better? 
I don't know how to read these figures so would be very grateful for an explanation.
Many thanks


Adrian Wolfson
DP, London
www,adrianwolfson.com

Geoff Boyle
 

Lower is better.

cheers
Geoff Boyle


From: cml-general@... <cml-general@...> on behalf of Adrian Wolfson <adrian@...>
Sent: Monday, February 5, 2018 6:03:21 PM
To: cml-general@...
Subject: Re: [general] Camera noise measurements
 
Hi Geoff
I would really like to understand what your test actually means.
Are you saying Varicam has lowest noise level or Alexa?
Is green gain lower better?
Is NR values lower or higher better? 
I don't know how to read these figures so would be very grateful for an explanation.
Many thanks


Adrian Wolfson
DP, London
www,adrianwolfson.com

Adrian Wolfson
 

Hi Geoff
I would really like to understand what your test actually means.
Are you saying Varicam has lowest noise level or Alexa?
Is green gain lower better?
Is NR values lower or higher better? 
I don't know how to read these figures so would be very grateful for an explanation.
Many thanks


Adrian Wolfson
DP, London
www,adrianwolfson.com

Alex Metcalfe DoP
 

No opinions, simple facts

Always nice when facts back up observation. Thanks Geoff that's useful
to be proved.


On 05/02/2018 16:52, Geoff Boyle wrote:
No opinions, simple facts
--
Alex Metcalfe DoP
00 44 (0) 7785 557611
www.alexmetcalfedop.co.uk

Geoff Boyle
 

A lot of opinions have been thrown in here recently so let’s look at some facts.

 

I took raw files into Resolve and using the 3 stop under exposures I altered offset only to match the backgrounds at “correct” level.

 

I then added simple noise reduction and adjusted it until no noise showed on the waveform.

 

The results are:-

 

Alexa SXT             Green Gain 44.46              NR 11.8

C700 4.5K            Green Gain 47.31              NR 18.8

F55                         Green Gain 50.50              NR 23.6

F65                         Green Gain 48.55              NR 21.2

Helium                  Green Gain 48.19              NR 19.0

Varicam Pure     Green Gain 48.02              NR 10.4

 

You can do this test yourself, it’s easy to download the EXR’s from the website.

 

No opinions, simple facts.

 

Cheers

 

Geoff Boyle

DP

Netherlands