Canon EOS Noise Reduction (NR) for Different Drive/Shutter Modes

Started 2 months ago | Discussions
bclaff Forum Pro • Posts: 10,474
Canon EOS Noise Reduction (NR) for Different Drive/Shutter Modes
8

Noise Reduction (NR), even with all menu settings OFF, is visualized using 2-Dimensional Fourier Transforms (2D FTs).
When no NR is present the pattern is random; the less random the more NR.

Mosaics are arranged in whole ISO setting order from ISO 50 to ISO 102400
NR is "expected" at 102400 but the NR ar low ISO is "unexpected".

Probably best viewed at "original size".

Mechanical Shutter

High Speed Continuous

Electronic

-- hide signature --

Bill ( Your trusted source for independent sensor data at PhotonsToPhotos )

(unknown member) Regular Member • Posts: 376
Any perceptual loss of resolution?

What are the downsides here? Is there any perceptual loss of resolution that you can detect?   Its a 45MP sensor and if it can have a bit more DR than a 42MP sensor would that be a good trade off, or is there more I am missing?

SHood Veteran Member • Posts: 5,624
Re: Canon EOS Noise Reduction (NR) for Different Drive/Shutter Modes

bclaff wrote:

Noise Reduction (NR), even with all menu settings OFF, is visualized using 2-Dimensional Fourier Transforms (2D FTs).
When no NR is present the pattern is random; the less random the more NR.

Mosaics are arranged in whole ISO setting order from ISO 50 to ISO 102400
NR is "expected" at 102400 but the NR ar low ISO is "unexpected".

Probably best viewed at "original size".

Mechanical Shutter

High Speed Continuous

Electronic

That is quite interesting.  Would this suggest that the lower DR shown in your other thread for the electronic shutter is mainly due to less applied NR in this mode?

-- hide signature --
 SHood's gear list:SHood's gear list
Olympus E-M1 II Olympus OM-D E-M1X Olympus M.Zuiko 300mm F4 IS Pro Olympus 12-100mm F4.0 Olympus 17mm F1.2 Pro
OP bclaff Forum Pro • Posts: 10,474
Re: Any perceptual loss of resolution?

Willow w wrote:

What are the downsides here? Is there any perceptual loss of resolution that you can detect? Its a 45MP sensor and if it can have a bit more DR than a 42MP sensor would that be a good trade off, or is there more I am missing?

It's impossible to know how much detail is lost to the NR.

-- hide signature --

Bill ( Your trusted source for independent sensor data at PhotonsToPhotos )

OP bclaff Forum Pro • Posts: 10,474
Re: Canon EOS Noise Reduction (NR) for Different Drive/Shutter Modes

SHood wrote:

...

That is quite interesting. Would this suggest that the lower DR shown in your other thread for the electronic shutter is mainly due to less applied NR in this mode?

No. Typically DR is lower because the higher readout rate has more noise.
In this case the NR just makes it harder to separate that effect out.

-- hide signature --

Bill ( Your trusted source for independent sensor data at PhotonsToPhotos )

(unknown member) Regular Member • Posts: 376
Possibly virtually none
2

bclaff wrote:

Willow w wrote:

What are the downsides here? Is there any perceptual loss of resolution that you can detect? Its a 45MP sensor and if it can have a bit more DR than a 42MP sensor would that be a good trade off, or is there more I am missing?

It's impossible to know how much detail is lost to the NR.

The best comparison I can find is the DPR studio scene, but the R5 uses a sharper lens than what is on the A7R4 and Z7. The detail at ISO 100 (RAW) with the R5 is very close to the A7R4. A $2000+ lens will do that.

I think too many other factors make a bigger difference if there is any with the NR. Even a little shake or lack of stabilization, or poor atmospheric conditions with a long focal length probably matter more (and that sharper lens).

Anyway, enough of my opinion, thank you for your hard work.

EthanCam Forum Member • Posts: 65
Re: Canon EOS Noise Reduction (NR) for Different Drive/Shutter Modes

bclaff wrote:

Noise Reduction (NR), even with all menu settings OFF, is visualized using 2-Dimensional Fourier Transforms (2D FTs).
When no NR is present the pattern is random; the less random the more NR.

Mosaics are arranged in whole ISO setting order from ISO 50 to ISO 102400
NR is "expected" at 102400 but the NR ar low ISO is "unexpected".

Probably best viewed at "original size".

Mechanical Shutter

High Speed Continuous

Electronic

Thank you very much for this!

Okay, this confirms the NR is applied indifferently of the shooting mode, and your DR measurements for the ES/HS modes seem to show that the NR is not "boosted" in those modes to increase the DR.

I suspect then NR to be applied at a relatively low-level of the signal processing chain, or at least it seems like a possibility.

Mark3794 Junior Member • Posts: 46
Re: Canon EOS Noise Reduction (NR) for Different Drive/Shutter Modes

So this random pattern happens only with noise reduction or it could possibly be caused by other "trickery" like using in a different way the dual pixel architecture, DGO, or some hardware noise reduction after the adc?

 Mark3794's gear list:Mark3794's gear list
Canon EOS M50 Canon EF 70-200mm F4L IS USM Canon EF 24-70mm F4L IS USM Canon EF 50mm F1.8 STM
OP bclaff Forum Pro • Posts: 10,474
Re: Canon EOS Noise Reduction (NR) for Different Drive/Shutter Modes
1

Mark3794 wrote:

So this random pattern happens only with noise reduction or it could possibly be caused by other "trickery" like using in a different way the dual pixel architecture, DGO, or some hardware noise reduction after the adc?

The non-random pattern indicates a 2-dimensional correlation between pixels and their neighbors. (Random indicates no correlation.)
This isn't what you'd see with any of the approaches you mention.
It is what you see when an algorithm runs over the sensor array values and potentially replaces a pixel with something computed from it's neighbors.

-- hide signature --

Bill ( Your trusted source for independent sensor data at PhotonsToPhotos )

Mark3794 Junior Member • Posts: 46
Re: Canon EOS Noise Reduction (NR) for Different Drive/Shutter Modes

Yeah sorry it was a typo i meant non-random. Thank you for explaining

 Mark3794's gear list:Mark3794's gear list
Canon EOS M50 Canon EF 70-200mm F4L IS USM Canon EF 24-70mm F4L IS USM Canon EF 50mm F1.8 STM
stoffer Regular Member • Posts: 275
Re: Canon EOS Noise Reduction (NR) for Different Drive/Shutter Modes

bclaff wrote:

Mark3794 wrote:

So this random pattern happens only with noise reduction or it could possibly be caused by other "trickery" like using in a different way the dual pixel architecture, DGO, or some hardware noise reduction after the adc?

The non-random pattern indicates a 2-dimensional correlation between pixels and their neighbors. (Random indicates no correlation.)
This isn't what you'd see with any of the approaches you mention.
It is what you see when an algorithm runs over the sensor array values and potentially replaces a pixel with something computed from it's neighbors.

This can’t just be something that evens out a signal difference between the two photo diodes within  the pixel, right?

-- hide signature --

Best regards, Stoffer
May the Light be with you!

 stoffer's gear list:stoffer's gear list
Sony a7R III Sony FE 35mm F1.4 Sony FE 24mm F1.4 GM Apple iPhone X +1 more
Ayoh Regular Member • Posts: 388
Re: Canon EOS Noise Reduction (NR) for Different Drive/Shutter Modes
1

bclaff wrote:

Noise Reduction (NR), even with all menu settings OFF, is visualized using 2-Dimensional Fourier Transforms (2D FTs).
When no NR is present the pattern is random; the less random the more NR.

Mosaics are arranged in whole ISO setting order from ISO 50 to ISO 102400
NR is "expected" at 102400 but the NR ar low ISO is "unexpected".

Probably best viewed at "original size".

Mechanical Shutter

High Speed Continuous

Electronic

Makes sense that there is less low ISO NR visible in the electronic shutter mode since the DR is reduced anyway to a lower readout bit rate. So either less NR is applied or is just not as visible due to reduced DR

Seems Canon was a bit too eager to match the rivals and resorted to built in NR

Ayoh Regular Member • Posts: 388
Re: Possibly virtually none

Willow w wrote:

bclaff wrote:

Willow w wrote:

What are the downsides here? Is there any perceptual loss of resolution that you can detect? Its a 45MP sensor and if it can have a bit more DR than a 42MP sensor would that be a good trade off, or is there more I am missing?

It's impossible to know how much detail is lost to the NR.

The best comparison I can find is the DPR studio scene, but the R5 uses a sharper lens than what is on the A7R4 and Z7. The detail at ISO 100 (RAW) with the R5 is very close to the A7R4. A $2000+ lens will do that.

I think too many other factors make a bigger difference if there is any with the NR. Even a little shake or lack of stabilization, or poor atmospheric conditions with a long focal length probably matter more (and that sharper lens).

Anyway, enough of my opinion, thank you for your hard work.

The NR might be just being applied in the deep shadows

OP bclaff Forum Pro • Posts: 10,474
Re: Canon EOS Noise Reduction (NR) for Different Drive/Shutter Modes
2

stoffer wrote:

bclaff wrote:

Mark3794 wrote:

So this random pattern happens only with noise reduction or it could possibly be caused by other "trickery" like using in a different way the dual pixel architecture, DGO, or some hardware noise reduction after the adc?

The non-random pattern indicates a 2-dimensional correlation between pixels and their neighbors. (Random indicates no correlation.)
This isn't what you'd see with any of the approaches you mention.
It is what you see when an algorithm runs over the sensor array values and potentially replaces a pixel with something computed from it's neighbors.

This can’t just be something that evens out a signal difference between the two photo diodes within the pixel, right?

No. If that showed at all it would probably show as horizontal only; this is equally horizontal and vertical.

-- hide signature --

Bill ( Your trusted source for independent sensor data at PhotonsToPhotos )

Keyboard shortcuts:
FForum MMy threads