Equivalence thread (including posts moved here from other threads)

DanS

Mu-43 All-Pro
Joined
Mar 8, 2016
Messages
1,760
Image level noise is a new concept to me. Can you explain what you mean by that?
What he's talking about is taking a 16mp image and a 24mp image, and then printing them out at the same size. or say resizing them to 4mp. The 24mp image will look better because you compressed(pick a verb) the image more.

It's extremely flawed logic in my opinion, because it breaks down when you aren't drastically resizing your image.

Imo, the only thing that matters is "Per-pixel" noise.
 

Speedliner

Mu-43 Hall of Famer
Joined
Mar 2, 2015
Messages
2,668
Location
Southern NJ, USA
Real Name
Rob
So cropping increases noise?

Fred
No not technically. But lets use an example. Let's say you have an image you like and you print it 8x10. Then you decide you don't quite like it as is and you go back to your original and crop it down a god bit. Now you print the cropped image out to 8x10 again. The cropped print will appear noisier because you magnified a smaller part of the original image to 8x10. Doing so magnifies the noise in the file along with the good parts.

It's hard to explain with words. Maybe there's a video on YouTube of the subject.
 

EarthQuake

Mu-43 Top Veteran
Joined
Sep 30, 2013
Messages
934
Image level noise is a new concept to me. Can you explain what you mean by that?

Fred
Certainly, by image level noise, I mean how much noise is visible when viewing an image in its entirety, like normal people view photos. As apposed to pixel-peeping per-pixel noise, which tells one very little about the overall quality of an image.

Here is an example. I've taken two photos of the same subject, at the same distance, with the shutter speed, ISO, and fstop, but at different focal lengths. One at 12mm, the other at 24mm. Then, I scaled down the 24mm to match the framing of the 12mm shot. Both photos were processed from raw in the same way.

Subscribe to see EXIF info for this image (if available)
Subscribe to see EXIF info for this image (if available)


When viewed at the same output size, the cropped image has visibly worse noise. The photosite size, of course, has not changed.

Consequently, what this shows is that, for equally efficient sensors, photosite size is not the determining factor for noise. The bigger sensor records more total light, more photons, which results in less noise.

This is a crude demonstration of the differences you would see with a 64MP 35mm sensor and 16MP 43 sensor, given perfectly equal performance between the sensors. In this case it's a 16MP 43 sensor vs a 4MP section of that same sensor, but the principal applies regardless of the specific sensor sizes or resolutions one would choose to compare.

Now, sensors are rarely perfectly equal when it comes to efficiency, but the variances in efficiency/performance tend to be much less than those between the basic physical limitations of a given sensor size.
 
Last edited:

DanS

Mu-43 All-Pro
Joined
Mar 8, 2016
Messages
1,760
Certainly, by image level noise, I mean how much noise is visible when viewing an image in its entirety, like normal people view photos. As apposed to pixel-peeping per-pixel noise, which tells one very little about the overall quality of an image.
Yes, because no one in the history of photography has ever had to aggressively crop an image, or enlarge it!:dash2:
 
Last edited:

pellicle

Mu-43 Hall of Famer
Joined
Feb 10, 2010
Messages
3,955
Location
Killarney, OzTrailEYa
Real Name
pellicle
So cropping increases noise?
no, increasing pixel density increases noise ... because there will be statistically fewer photons to be counted by each photo-site (thus increasing noise by reducing signal)

I'm not sure what you mean by reconfiguring.
I mean that instead of simply slicing a 5D sensor and making a 43rds sensor of the same density of pixels (which would be about a 5.46MegaPix 43rds sensor using the same pixel density as the 5DMkii data above), that you reconfigure that sliced sensor to be 16Mpix (which is what people see when they photograph with their 43rds and pixel peep). Its that that leads to more noise.

I was attempting to agree with you. I suspect if we had a 6Mpix 43rds camera (with essentially the same size photo-sites as the 5D has that it too would be exceptionally low noise (but of course not be as many pixels which matters for marketing and those who aggressively crop rather than use their lenses appropriately)
 
Last edited:

pellicle

Mu-43 Hall of Famer
Joined
Feb 10, 2010
Messages
3,955
Location
Killarney, OzTrailEYa
Real Name
pellicle
Yes, because no one in the history of photography has ever had to aggressively crop an image, or enlarge it!:dash2:
at the risk of feeding this (as everyone seems to be describing eggs with different words) I'll say that even before agressively cropping an image that 43rds is already a x2 crop of a full frame. However at manufacture that "crop" is also stuffed with more pixels resulting in each photo site getting less light

:)
 

pellicle

Mu-43 Hall of Famer
Joined
Feb 10, 2010
Messages
3,955
Location
Killarney, OzTrailEYa
Real Name
pellicle
...
Here is an example. I've taken two photos of the same subject, at the same distance, with the shutter speed, ISO, and fstop, but at different focal lengths. One at 12mm, the other at 24mm. Then, I scaled down the 24mm to match the framing of the 12mm shot. Both photos were processed from raw in the same way..
I think this is about as good an example as one could get without actually having a FF sensor with similar electrical properties as the 43rds sensor with both having the same pixel density.

I suspect with some good intelligent pixel binning (as perhaps done by Nokia on their pureview stuff) may even reduce the noise more on the "scaled down" image
 

pdk42

One of the "Eh?" team
Joined
Jan 11, 2013
Messages
6,631
Location
Leamington Spa, UK
Everything up to here is perfectly accurate and valid.



This is absolutely incorrect. By this logic, a 43 sensor with larger photosites than a corresponding 35mm sensor would produce an image with less noise, which of course, is unequivocally false.

FF sensors perform better because they have a larger area, which collects more total photons regardless of photosite size. If the FF sensor has a 12MP or 42MP sensor makes very little difference to the quality or noise level of the final image assuming the output is equalized (that is to say, printed and viewed at the same size and distance). See Sony A7s II and A7r II for a real world example, the noise performance between these cameras is very similar outside of extremely high ISO where the A7s has slight advantage.

Another example, the A7r II vs a 12MP 43 sensor, the photosites are close in size but the A7r produces an image with much less noise. This isn't entirely fair, as there is no modern 12MP 43 sensor, but even the recent 16MP 43 sensors only have photosites a bit smaller, but the A7r handily outperforms those sensors.

Per unit (such as mm squared) light gathering, and per-pixel noise performance are largely irrelevant outside of academic discussions for the sake of academics.
I think you missed my point. There are two reasons why FF sensors are better than u43 sensors:

- The photosites are bigger with correspondingly larger charge-well depth. This improves the per-pixel noise
- Given no cropping, you enlarge the FF image less. This reduces the apparent noise in the final output

I think we agree, it's just a question of how it's phrased. I personally hate this "total amount of light/more photons" line of reasoning since all my instincts tell me we shuold be looking at the light per unit area.
 

pellicle

Mu-43 Hall of Famer
Joined
Feb 10, 2010
Messages
3,955
Location
Killarney, OzTrailEYa
Real Name
pellicle
...I personally hate this "total amount of light/more photons" line of reasoning since all my instincts tell me we shuold be looking at the light per unit area.
I think it comes from people who grew up reading text books in the film era who's authors ignored or oversimplified the fact that photons still had to interact with halides.

In digital there is a well that has a photon counter at the bottom. Same thing really.

Looked at another way if you put out a bucket or a test tube in the rain both will catch the rain. If the rain is sparse then the bucket is more likely to be accurate than a thin test tube in registering how many inches of rain fell.

You won't find a rain gauge in fractions of an inch opening for that reason.

A good primer
Does Pixel Size Matter Clarkvision.com
 

fredlong

Just this guy...
Joined
Apr 18, 2011
Messages
1,753
Location
Massachusetts USA
Real Name
Fred
Certainly, by image level noise, I mean how much noise is visible when viewing an image in its entirety, like normal people view photos. As apposed to pixel-peeping per-pixel noise, which tells one very little about the overall quality of an image.

Here is an example. I've taken two photos of the same subject, at the same distance, with the shutter speed, ISO, and fstop, but at different focal lengths. One at 12mm, the other at 24mm. Then, I scaled down the 24mm to match the framing of the 12mm shot. Both photos were processed from raw in the same way.

Subscribe to see EXIF info for this image (if available)
Subscribe to see EXIF info for this image (if available)


When viewed at the same output size, the cropped image has visibly worse noise. The photosite size, of course, has not changed.

Consequently, what this shows is that, for equally efficient sensors, photosite size is not the determining factor for noise. The bigger sensor records more total light, more photons, which results in less noise.

This is a crude demonstration of the differences you would see with a 64MP 35mm sensor and 16MP 43 sensor, given perfectly equal performance between the sensors. In this case it's a 16MP 43 sensor vs a 4MP section of that same sensor, but the principal applies regardless of the specific sensor sizes or resolutions one would choose to compare.

Now, sensors are rarely perfectly equal when it comes to efficiency, but the variances in efficiency/performance tend to be much less than those between the basic physical limitations of a given sensor size.
Ok, so bigger noise equals more noise. I disagree. Signal to noise ratio is the same.

And if you crop out three quarters of the noise, don't you actually have less overall noise ;)
 

ijm5012

Mu-43 Legend
Joined
Oct 2, 2013
Messages
7,960
Location
Pittsburgh, PA
Real Name
Ian
No not technically. But lets use an example. Let's say you have an image you like and you print it 8x10. Then you decide you don't quite like it as is and you go back to your original and crop it down a god bit. Now you print the cropped image out to 8x10 again. The cropped print will appear noisier because you magnified a smaller part of the original image to 8x10. Doing so magnifies the noise in the file along with the good parts.

It's hard to explain with words. Maybe there's a video on YouTube of the subject.
Why not just frame your shot better in the first place, or use the appropriate lens? I don't go trying to shoot motorsports with my PL25...
 

DanS

Mu-43 All-Pro
Joined
Mar 8, 2016
Messages
1,760
Blah blah equivalence, noise, sensor size, etc. So over this cr@p
Understanding how and why stuff works, is more important today, than at any point in history. It baffles my mind, that the average consumer basically wants to be blissfully unaware nowadays.
 

Turbofrog

Mu-43 Legend
Joined
Mar 21, 2014
Messages
5,362
huh? Pretty much all image editing software has a crop feature.
Speedliner was saying that the image quality effects of printing the original image and then cropping the image and printing it at the same size is analogous to taking the image with a bigger sensor camera and a smaller sensor camera and then printing them both at the same size.

It's a reasonable statement in my mind.

Anyway, EarthQuake showed it perfectly in a visual way with his alarm clock photos. You can't really get clearer than that by way of explanation.
 
Links on this page may be to our affiliates. Sales through affiliate links may benefit this site.
Mu-43 is a fan site and not associated with Olympus, Panasonic, or other manufacturers mentioned on this site.
Forum post reactions by Twemoji: https://github.com/twitter/twemoji
Copyright © 2009-2019 Amin Forums, LLC
Top Bottom