Ugly Hedgehog - Photography Forum
Home Active Topics Newest Pictures Search Login Register ☰
Main Photography Discussion
Why does cropping reduce dynamic range?
Page <<first <prev 8 of 10 next> last>>
Feb 26, 2021 11:03:56   #
bleirer
 
bclaff wrote:
The original post:

The quick answer:
Since you mention PhotonsToPhotos to are talking about Photographic Dynamic Range (PDR).
PDR is measured using a standard final image size and viewing distance (not unlike Depth of Field (DOF))
PDR also uses a Signal to Noise Ratio (SNR) criteria (unlike DxOMark which uses only read noise).
Since the DX image has to be enlarged more than the FX image the SNR is lower resulting in a lower PDR.
Pixel size does not matter.
Also, remember, we're talking about the capability of the sensor and not some specific image of a scene with a particular dynamic range (the horsepower of your car is the same travelling at 10 mph than it is at 55 mph you're just using less of it).
The original post: br br The quick answer: br Sin... (show quote)


In your view can I take it as a general principal that cropping an image, any image from any camera, will result in a loss of dynamic range compared to not cropping?

And would the loss be proportional to ____?

Or are there any generalizations one could carry into the field?

Reply
Feb 26, 2021 11:17:38   #
selmslie Loc: Fernandina Beach, FL, USA
 
bclaff wrote:
Not true. At DxOMark their dynamic range is based entirely on something called (pixel) read noise and "normalized" for pixel size (resolution).
The SNR 18% charts have nothing to do with their dynamic range.

They only show the DR for full frame but they have two of them, a Screen version and a Print version which is about a full stop higher. But it really doesn't matter what they base it on. It's up to the viewer to compare the maximum to their arbitrary 6 Ev noise level and draw their own conclusions.

Most of us only used to compare one model against another. Their numbers are not all that credible anyhow.
bclaff wrote:
Except that it's not in proportion to the crop factor.

It's pretty close to about 1 stop (off by less thab 0.1 stops) for a DX crop on the D850, D610, Df and Z7. I'd say that's a pretty close correlation.

But once you crop the file on your computer the actual proportions are no longer important. The noise gets easier to see but nobody is going to measure that. It's either tolerable, fixable or too much.

Reply
Feb 26, 2021 11:22:09   #
nathanweddings
 
Huh?

Reply
Check out Panorama section of our forum.
Feb 26, 2021 11:33:17   #
selmslie Loc: Fernandina Beach, FL, USA
 
bclaff wrote:
That is only a rough approximation. The details as to why not are in the article I cited earlier ...

For example, for the Nikon D850 the drop in Photographic Dynamic Range (PDR) from FX to DX is about 3/4 stops while your reasoning would imply something closer to 7/6 stops.

Take a look at what you published for the D850:

FX DX Stops
ISO 800 8.86 7.81 1.05
ISO 6400 5.93 4.82 1.11
ISO 25600 3.89 2.84 1.05

I didn't look at all ISO levels but the difference is pretty close until you get down to low ISO settings.

Reply
Feb 26, 2021 13:19:18   #
bclaff Loc: Sherborn, MA (18mi SW of Boston)
 
selmslie wrote:
Take a look at what you published for the D850:

FX DX Stops
ISO 800 8.86 7.81 1.05
ISO 6400 5.93 4.82 1.11
ISO 25600 3.89 2.84 1.05

I didn't look at all ISO levels but the difference is pretty close until you get down to low ISO settings.

The low ISO settings are important and there the difference is as I stated.
For example at ISO 64 11.63 for FX and 10.90 for DX is a difference of 0.73

Reply
Feb 26, 2021 13:20:31   #
bclaff Loc: Sherborn, MA (18mi SW of Boston)
 
bleirer wrote:
In your view can I take it as a general principal that cropping an image, any image from any camera, will result in a loss of dynamic range compared to not cropping?

And would the loss be proportional to ____?

Or are there any generalizations one could carry into the field?

Roughly but not strictly proportional to the area used.

Reply
Feb 26, 2021 13:30:32   #
bclaff Loc: Sherborn, MA (18mi SW of Boston)
 
bclaff wrote:

Not true. At DxOMark their dynamic range is based entirely on something called (pixel) read noise and "normalized" for pixel size (resolution).
selmslie wrote:
They only show the DR for full frame but they have two of them, a Screen version and a Print version which is about a full stop higher. But it really doesn't matter what they base it on.
Screen is pixel level read noise and that is what I call Engineering Dynamic Range (EDR). Print is simply the Screen value "normalized" for pixel size. Screen has nothing to do with DX versus FX. DxOMark has no information for crop modes; period.
selmslie wrote:
It's pretty close to about 1 stop (off by less than 0.1 stops) for a DX crop on the D850, D610, Df and Z7. I'd say that's a pretty close correlation.
DxOMark publishes nothing that is relevant for crop modes; period. Any correlation you see is coincidental.
selmslie wrote:
But once you crop the file on your computer the actual proportions are no longer important. The noise gets easier to see but nobody is going to measure that. It's either tolerable, fixable or too much.
Well, actually, I do predict that for cropping DX from FX, and it is helpful to know when comparing.

Reply
Check out True Macro-Photography Forum section of our forum.
Feb 26, 2021 13:32:45   #
Drbobcameraguy Loc: Eaton Ohio
 
Ysarex wrote:
Noise increases because of the different total areas of the sensor used. Use the whole sensor and you record more total light than if you use the smaller cropped area. Noise will increase/decrease with the total amount of light recorded. SNR is a ratio and goes up or down with the total amount of light recorded.

Try this analogy: You have two cookie pans -- one is 11 x 15 inches and the other is 9 x 12 inches. You put them both out in the rain together and allow them to collect water for 1 minute. That 1 minute is exposure time. After a minute it stops raining and you measure the depth of water in each pan -- it's the same. That's exposure. But now pour the water from both pans into two separate bowls. Do both bowls have the same amount of water? The larger pan collected more water. In the same way a larger sensor will collect more light. Noise (shot noise which is the noise we're concerned with) as a ratio to the signal will decrease with more light collected -- stronger signal.

All else the same, bigger sensors are less noisy than smaller sensors. Noise limits DR.
Noise increases because of the different total are... (show quote)


Excellent way to explain it.

Reply
Feb 26, 2021 13:33:14   #
selmslie Loc: Fernandina Beach, FL, USA
 
bclaff wrote:
The low ISO settings are important and there the difference is as I stated.
For example at ISO 64 11.63 for FX and 10.90 for DX is a difference of 0.73

Below ISO 800 there is hardly enough noise for anyone to notice unless they underexpose. This is also an ISO level where the dynamic range plots start to get a bit strange. We may not care that the loss is only about 0.7 stops if we can't see noise in either case.

The high ISO settings are a more consistent indicator of the loss from cropping and it's close to 1 stop for most models. That's also where we are more concerned about noise.

Reply
Feb 26, 2021 13:46:55   #
selmslie Loc: Fernandina Beach, FL, USA
 
bclaff wrote:
Well, actually, I do predict that for cropping DX from FX, and it is helpful to know when comparing.

While we appreciate having the information we might be hard pressed to figure out when to apply it.

Anyone who uses a DX crop on an FX sensor is probably trying to make up for the fact that they are not using a lens with enough reach. They could just as easily shoot at FX and crop later. But a DX crop is fixed at 1.5x for Nikon and that cuts the megapixels down by 1/2.25 (it's worse for Canon).

Shooting at FX gives them more flexibility. They will then see while they are cropping whether the noise increases enough from cropping to make a difference.

Reply
Feb 26, 2021 13:50:08   #
Bridges Loc: Memphis, Charleston SC, now Nazareth PA
 
Ysarex wrote:
Noise increases because of the different total areas of the sensor used. Use the whole sensor and you record more total light than if you use the smaller cropped area. Noise will increase/decrease with the total amount of light recorded. SNR is a ratio and goes up or down with the total amount of light recorded.

Try this analogy: You have two cookie pans -- one is 11 x 15 inches and the other is 9 x 12 inches. You put them both out in the rain together and allow them to collect water for 1 minute. That 1 minute is exposure time. After a minute it stops raining and you measure the depth of water in each pan -- it's the same. That's exposure. But now pour the water from both pans into two separate bowls. Do both bowls have the same amount of water? The larger pan collected more water. In the same way a larger sensor will collect more light. Noise (shot noise which is the noise we're concerned with) as a ratio to the signal will decrease with more light collected -- stronger signal.

All else the same, bigger sensors are less noisy than smaller sensors. Noise limits DR.
Noise increases because of the different total are... (show quote)


Great analogy! You should be a teacher!

Reply
Check out Bridge Camera Show Case section of our forum.
Feb 26, 2021 14:10:11   #
Ysarex Loc: St. Louis
 
Bridges wrote:
Great analogy! You should be a teacher!


Thanks -- glad it made sense for you.

Reply
Feb 26, 2021 15:53:56   #
selmslie Loc: Fernandina Beach, FL, USA
 
selmslie wrote:
It's pretty close to about 1 stop (off by less than 0.1 stops) for a DX crop on the D850, D610, Df and Z7. I'd say that's a pretty close correlation.
bclaff wrote:
DxOMark publishes nothing that is relevant for crop modes; period. Any correlation you see is coincidental.

I found the correlation at Photons to Pixels for ISO settings above 400.

Obviously not at DxOMark because they don'd show FX vs. DX dynamic range plots.

Reply
Feb 26, 2021 16:01:27   #
Canisdirus
 
selmslie wrote:
But DR is still measured from the noise up, regardless of the pixel size.


It all stems from the pixels...the very first things to act upon the image coming from the lens.
ISO...DR...focus...all from the pixel setup and generation...and size.

As to someone mentioning the same image being cropped. with the same sensor...
That's because there is less image...less light..less pixels...which means less dynamic range.
A cropped image is not the same as the whole image. It's all processed by how much light there is to work with.

I have the A7RIV...shoots at 61MP (depending on the data of image)...
But I can crop instantly with a button...same sensor.
Now, however, I am shooting at 26MP....with less light... fewer pixels to gather light and information.
Dynamic range instantly falls off.

Reply
Feb 26, 2021 16:16:49   #
selmslie Loc: Fernandina Beach, FL, USA
 
Canisdirus wrote:
... As to someone mentioning the same image being cropped. with the same sensor...
That's because ...

That's whole reason for this thread as you can see in the first post. https://www.uglyhedgehog.com/t-687091-1.html#12030769

The fact that it happens is clear in the plots from Photons to Photos. Your camera loses nearly a full stop of DR at ISO 400 when you switch to crop mode (see attached). It's a little different at other ISO settings.

The reason it happens is primarily that the cropped image needs more enlarging to make the same size image as the one from the full sensor.


(Download)

Reply
Page <<first <prev 8 of 10 next> last>>
If you want to reply, then register here. Registration is free and your account is created instantly, so you can post right away.
Check out Software and Computer Support for Photographers section of our forum.
Main Photography Discussion
Home | Latest Digest | Back to Top | All Sections
Contact us | Privacy policy | Terms of use | DMCA
UglyHedgehog.com - Forum
Copyright 2011-2024 Ugly Hedgehog, Inc.