Pages: 1 2 3 [4] 5   Go Down

Author Topic: Lightroom 12.3 - Denoise is there  (Read 6708 times)

NikoJorj

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1082
    • http://nikojorj.free.fr/
Re: Lightroom 12.3 - Denoise is there
« Reply #60 on: May 20, 2023, 05:08:07 am »

The term "ISO invariance" refers to a camera's ability to maintain consistent image quality while capturing images at different ISO settings.
Just to nitpick a bit and add to my former message, this does only concern the shadows.
In the highlights, higher ISO may mean more clipped highlight (and therefore a loss of information).

The term that I'd prefer for cameras with good ISO behaviour is ISOless, because you don't need to set a compromise between highlight clipping and read noise in the shadows, and the ISO setting just stay put at the base (optimum) value.

Guillermo Luijk spoke of that much better than I could, see eg https://forum.luminous-landscape.com/index.php?topic=56906.0
Logged
Nicolas from Grenoble
A small gallery

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Lightroom 12.3 - Denoise is there
« Reply #61 on: May 20, 2023, 09:39:43 am »

Just to nitpick a bit and add to my former message, this does only concern the shadows.
In the highlights, higher ISO may mean more clipped highlight.
Due to that expose yes?
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

NikoJorj

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1082
    • http://nikojorj.free.fr/
Re: Lightroom 12.3 - Denoise is there
« Reply #62 on: May 20, 2023, 04:29:41 pm »

Due to that expose yes?
Clipping due to ISO I'd rather say : with constant exposure, a higher ISO will cut more of the highlights due to the gain applied.
Logged
Nicolas from Grenoble
A small gallery

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Lightroom 12.3 - Denoise is there
« Reply #63 on: May 20, 2023, 04:31:35 pm »

Clipping due to ISO I'd rather say : with constant exposure, a higher ISO will cut more of the highlights due to the gain applied.
Got it; thanks.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Simon J.A. Simpson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 584
Re: Lightroom 12.3 - Denoise is there
« Reply #64 on: May 21, 2023, 04:46:08 pm »

Clipping due to ISO I'd rather say : with constant exposure, a higher ISO will cut more of the highlights due to the gain applied.

I don't think this is correct.

I've just made a test; two sets of exposures of an X-Rite Colorchecker, one set at ISO 200 and one set at ISO 6400, successively increasing the exposure to +5 stops.  I can see no visible difference between the ISO 200 clipping and ISO 6400 clipping at the same levels of over exposure.

This makes sense since at higher ISOs the light meter is recommending successively reduced exposures (less light on the sensor) therefore resulting in a reduced signal which is then being amplified before going to the ADC (Analogue to Digital Converter).  So, I would expect more noise but I would not expect highlight clipping at higher ISOs.

As regards to increasing ISO and noise, I believe this is more complicated with more modern cameras.  Initial testing of my new DSLR suggests that there may be some kind of noise reduction being applied; but, as a general rule, the higher the ISO the more noise I am getting, particularly in the shadows – although I have to say the high ISO results I am achieving are astonishingly good compared to my previous camera.

I am not sure I subscribe to the "ISO-invariance" idea.  I don't think that any digital sensor/camera can actually be ISO-invariant although it might be possible to describe them as 'noise-invariant'; or perhaps, more accurately, as 'less noise-variant'.
« Last Edit: May 21, 2023, 04:53:09 pm by Simon J.A. Simpson »
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Lightroom 12.3 - Denoise is there
« Reply #65 on: May 21, 2023, 05:09:52 pm »

What camera Simon?
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

fdisilvestro

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1852
    • Frank Disilvestro
Re: Lightroom 12.3 - Denoise is there
« Reply #66 on: May 21, 2023, 08:02:35 pm »

I don't think this is correct.

I've just made a test; two sets of exposures of an X-Rite Colorchecker, one set at ISO 200 and one set at ISO 6400, successively increasing the exposure to +5 stops.  I can see no visible difference between the ISO 200 clipping and ISO 6400 clipping at the same levels of over exposure.

This makes sense since at higher ISOs the light meter is recommending successively reduced exposures (less light on the sensor) therefore resulting in a reduced signal which is then being amplified before going to the ADC (Analogue to Digital Converter).  So, I would expect more noise but I would not expect highlight clipping at higher ISOs.



You are changing exposure. The previous comment was about keeping the same exposure (shutter speed / aperture) while changing ISO.

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Lightroom 12.3 - Denoise is there
« Reply #67 on: May 21, 2023, 08:14:05 pm »

You are changing exposure. The previous comment was about keeping the same exposure (shutter speed / aperture) while changing ISO.
Indeed.
Now this is an old Canon DSLR, and as mentioned, it isn't ISO Invariant.
Same Exposure, only difference is the ISO (100 vs. 800).
I've placed the eyedropper over the same area in Lightroom Classic and set the readout for Lab. Unfortunately, it's nearly impossible in that product to control the exact position of the eyedropper, unlike Photoshop or the sampling. But you can see the differences are tiny and insignificant. It appears in this case, the ISO does not affect the highlights. YMMV.

Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Simon J.A. Simpson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 584
Re: Lightroom 12.3 - Denoise is there
« Reply #68 on: May 22, 2023, 05:12:37 am »

You are changing exposure. The previous comment was about keeping the same exposure (shutter speed / aperture) while changing ISO.

Yes, a good point.

But if you set the ISO to a higher figure and keep the exposure the same then the resulting file will have higher tonal values thus clipping the highlights.  By way of example, if you expose a grey card at 200 ISO to produce a mid-grey (nominally, the correct exposure) and then increase the ISO to 800, say, but keep the exposure the same, then you have effectively pushed the grey-card exposure into the highlights at +3 stops.  So, I would expect to start to see some clipping.

However, I think we may be missing something in our understanding.

THE SENSOR
In analogue terms, before the signal is sent to the ADC, setting a higher ISO would in theory give the sensor (before the signal is amplified) more headroom since the exposure, as advised by the meter, is reduced.  So, in theory, increasing the ISO by 5 stops, as in my test, would give an extra 5 stops more headroom above mid-grey (assuming that the ‘base’ ISO is 200).  But my results show identical tonal compression and loss of colour fidelity (clipping) at the same amounts of over-exposure for both ISO 200 and 6400.

At the moment I don’t really know why this should be.  It could be down to the ADC or some other digital processing.  I need to think about this some more and perhaps do some more tests.

As I understand it, at the moment, on the analogue side of things the sensor is limited by the maximum amount of volts it can produce.  Exposing it to more light will not increase its output voltage (full bucket).  This varying voltage or signal is produced when the sensor is ‘read’ and all those buckets full of electrons are emptied  ;D.

SETTING ISO
I believe the ISO setting on the camera, or more properly the Exposure Index (E.I.), does at least three things.  It changes:
   •   the sensitivity of the camera’s exposure meter
   •   the analogue amplification the signal sent to the ADC
   •   the ISO flag in the EXIF data to the ISO selected.
I suspect that it may do more than this.  Possibly it may introduce analogue noise reduction at higher ISOs or introduce some processing of the resulting digital RAW file.  Current evidence from some modern DSLRs and higher-end mirrorless cameras suggest this may be the case.
Logged

Simon J.A. Simpson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 584
Re: Lightroom 12.3 - Denoise is there
« Reply #69 on: May 22, 2023, 05:15:48 am »

What camera Simon?

I was being deliberately coy about the camera I tested as I didn’t want to set-off a fan-boy type exchange with other contributors.

The camera I am testing is a Canon EOS 5D MkIV.  Previously, I have run tests on the 5D and the 5D MkIII.  I do not currently have access to other cameras but the Nikon D8XX series reportedly has outstanding noise performance.  I do not know much about the Sonys, although I understand they share the same sensor as the Nikons (and are spoken of very highly by their owners).  I am very interested in the results from your Canon R6.

My own interest is in trying to understand how sensors work, how ISO works, and the dynamic range of the sensor of my camera, so that I can better determine my exposures for the best results.  This stems from a lifelong interest in photography à la the Ansel Adams ‘school’ of photography.  If you’ve read Ansel Adams book ‘The Negative’ you will realise my tests replicate those he recommended for understanding how film works.
Logged

fdisilvestro

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1852
    • Frank Disilvestro
Re: Lightroom 12.3 - Denoise is there
« Reply #70 on: May 22, 2023, 06:33:47 am »


THE SENSOR
In analogue terms, before the signal is sent to the ADC, setting a higher ISO would in theory give the sensor (before the signal is amplified) more headroom since the exposure, as advised by the meter, is reduced.  So, in theory, increasing the ISO by 5 stops, as in my test, would give an extra 5 stops more headroom above mid-grey (assuming that the ‘base’ ISO is 200).  But my results show identical tonal compression and loss of colour fidelity (clipping) at the same amounts of over-exposure for both ISO 200 and 6400.


You cannot use that "extra" headroom. The analogue gain is applied to the signal before the Analog to digital conversion (ADC), so the ADC reaches saturation at a lower signal than compared to a lower gain. There is nothing mysterious here.


Simon J.A. Simpson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 584
Re: Lightroom 12.3 - Denoise is there
« Reply #71 on: May 22, 2023, 08:53:26 am »

You cannot use that "extra" headroom. The analogue gain is applied to the signal before the Analog to digital conversion (ADC), so the ADC reaches saturation at a lower signal than compared to a lower gain. There is nothing mysterious here.

So that would suggest that the clipping (tonal compression and loss of colour fidelity) is a function of the ADC rather than the sensor itself ?

The voltages from the sensor, prior to being amplified, are directly proportional to the amount of light falling on it.  So, if the sensor saturates or clips at +3 stops and then you under expose by 5 stops (prior to amplification) the sensor now has 8 stops before it saturates or clips.  No?

I'm not, as yet, convinced since the ADC is simply creating binary encoding of the voltages from the output of the amplifier.  I can understand how the voltages, being analogue, might vary according to the output of the sensor (then amplified), but not the resulting binary; unless something else is happening during conversion ?  Possibly?

More thought needed…
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Lightroom 12.3 - Denoise is there
« Reply #72 on: May 22, 2023, 11:27:14 am »

I was being deliberately coy about the camera I tested as I didn’t want to set-off a fan-boy type exchange with other contributors.
But the kind/make can make a difference here.
Out Canon's are not ISO Invariant.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

fdisilvestro

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1852
    • Frank Disilvestro
Re: Lightroom 12.3 - Denoise is there
« Reply #73 on: May 22, 2023, 03:49:17 pm »

So that would suggest that the clipping (tonal compression and loss of colour fidelity) is a function of the ADC rather than the sensor itself ?

The voltages from the sensor, prior to being amplified, are directly proportional to the amount of light falling on it.  So, if the sensor saturates or clips at +3 stops and then you under expose by 5 stops (prior to amplification) the sensor now has 8 stops before it saturates or clips.  No?

I'm not, as yet, convinced since the ADC is simply creating binary encoding of the voltages from the output of the amplifier.  I can understand how the voltages, being analogue, might vary according to the output of the sensor (then amplified), but not the resulting binary; unless something else is happening during conversion ?  Possibly?

More thought needed…

Suppose you have an 8 bit ADC to convert a signal from 0 to 1V, then an input of 1V will result in the maximum number out of the encoder, which is 255.

The ADC does not know if that 1V is because the signal out of the sensor was 1V or if it was 0.2V with a gain of 5x. It is the ADC that reaches the maximum encoding value when the "Clipping" occurs.

As far as the ADC is concerned, the signal out of the sensor multiplied by the gain is a black box

Simon J.A. Simpson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 584
Re: Lightroom 12.3 - Denoise is there
« Reply #74 on: May 22, 2023, 04:03:02 pm »

Suppose you have an 8 bit ADC to convert a signal from 0 to 1V, then an input of 1V will result in the maximum number out of the encoder, which is 255.

The ADC does not know if that 1V is because the signal out of the sensor was 1V or if it was 0.2V with a gain of 5x. It is the ADC that reaches the maximum encoding value when the "Clipping" occurs.

As far as the ADC is concerned, the signal out of the sensor multiplied by the gain is a black box

Yes.  I agree.

But the clipping which my tests show is that with increased exposure first there is 'soft' clipping where tonal separation is compressed and there is a gradual loss of colour fidelity as exposure is increased.  This can also be observed in the graphs from RawDigger.

ADC clipping will be 'hard'; a complete loss of variation in the signal; i.e. highlights with no discernible nor retrievable detail.

This makes me wonder whether the 'soft' clipping is the result of the signal coming from the analogue sensor.  Or is this being introduced by some digital processing ?  I don’t think it would occur in the situation you postulate as this would result in completely blown highlights, nothing there at all.

Puzzling…

And worth thinking about some more.
Logged

fdisilvestro

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1852
    • Frank Disilvestro
Re: Lightroom 12.3 - Denoise is there
« Reply #75 on: May 22, 2023, 05:24:17 pm »

My example above was an oversimplification. Many things may be at play there. The signal may be "compressed" to avoid hard clipping, but this is only a possibility.

In any case, my point is that the "extra" headroom in the sensor is not available to use, at least not keeping a linear response of the whole system (sensor + ADC)

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Lightroom 12.3 - Denoise is there
« Reply #76 on: May 22, 2023, 06:42:48 pm »

Hard Clipping?
Soft Clipping?
OK, I'm game what's what here?
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

fdisilvestro

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1852
    • Frank Disilvestro
Re: Lightroom 12.3 - Denoise is there
« Reply #77 on: May 22, 2023, 07:29:16 pm »

Hard clipping -> the output increases linearly until it reached the maximum value of the encoder and there is an abrupt transition. Typical response of basic, uncorrected digital systems.

In soft clipping, the output "shoulder-off" and there is a gradual transition until saturation or maximum value. This response is more common in analogue systems and more pleasant to the observer.

The use of compression of the analogue signal, both in audio and images, allow for soft clipping at the expense of lower fidelity to the original signal.



digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Lightroom 12.3 - Denoise is there
« Reply #78 on: May 22, 2023, 08:18:40 pm »

Hard clipping -> the output increases linearly until it reached the maximum value of the encoder and there is an abrupt transition. Typical response of basic, uncorrected digital systems.

In soft clipping, the output "shoulder-off" and there is a gradual transition until saturation or maximum value. This response is more common in analogue systems and more pleasant to the observer.

The use of compression of the analogue signal, both in audio and images, allow for soft clipping at the expense of lower fidelity to the original signal.
I'd like to see a photo since I'm still a bit confused.
In each case, there's a point of complete clipping right?
Maybe the term should be smooth clipping and abrupt clipping?
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

fdisilvestro

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1852
    • Frank Disilvestro
Re: Lightroom 12.3 - Denoise is there
« Reply #79 on: May 22, 2023, 08:25:30 pm »


Maybe the term should be smooth clipping and abrupt clipping?

Yes, it could be that!
Pages: 1 2 3 [4] 5   Go Up