A technical question with regards to down-sizing

[Hugoneus]

Hi,

If have two images, one 12MP and one 36MP and they both appear to have the same noise when viewed at 100%, how much noise will I eliminate if I down-size the 36MP image to 12MP and view it at 100%? If I now put both these 12MP images side by side, will the down-sized image look much more clean?

Thanks!

 

[Hugoneus]

Amazing how rumor posts get 100s of replies, but a legitimate question... nothing.

 

[rhlpetrus]

Think of 36 MP as a 3x larger image size than 12MP in area (which is the case using same resolution output device). In terms of stops that means about 1.7 stop in favor or larger image, thus reducing it to 12MP, in principle, would gain 1.7 stop in all areas, in terms of perception at least.

Now, if you are thinking in terms of the D7000 sensor upsized to FF, it would be 36MP. By most tests and reviews, in high ISO it is about 1 stop under the D700 and about 1.7 stop under D3s. Using DxO mark, for example, cut points for high ISO in noise terms are:

D7000: ISO 1167
D700: ISO 2303
D3s: ISO 3253

THus, resizing a 36MP file with same performance as the D7000's to 12MP would provide similar noise at 3200 ISO as the D3s.
--
Renato.
http://www.flickr.com/photos/rhlpedrosa/
OnExposure member
http://www.onexposure.net/

Good shooting and good luck
(after Ed Murrow)

 

[tarnish]

If have two images, one 12MP and one 36MP and they both appear to have the same noise when viewed at 100%,

The first trick will be in obtaining those two images. One way is to make them synthetically. Create two blank images of those sizes then use an Add Noise function in something like Photoshop.

A useful exercise would be to view those two images side by side at the same overall frame scale which should also be large enough to allow you to see the noise in the 12MP image.

This experiment approximately models the difference between 12MP DX and 36MP FX, assuming they were made with identical technology including the full signal handling path, and with very similar pixel scales. I say "approximate" because the area ratio between those formats doesn't happen to be 3X.

It's not a good model for the difference between 12MP FX and 36MP FX because of the difference in the count of captured photons per pixel . That will violate your condition of having the same noise appearance at 100%.

 

[Steve Bingham]

Let's look at it another way. If you take a 12mp picture and uprez (resample) it to a larger size you will introduce more noise - if the print size is large enough to see it! Doing the opposite should reduce the noise. However, I don't think the noise reduction will be as much as you might expect. Having no files from a 36mp camera available, it is difficult for me to test. And . . . I always prefer testing verses theory.

As all pixels are not created equal this then becomes non answerable. However, if the pixels have identical qualities then yes, noise will be reduced.

Can you be more specific as to which sensors you are talking about? There are people on this forum that know far more about sensors than I do!
--
Steve Bingham
http://www.dustylens.com
http://www.ghost-town-photography.com

 

[Steve Bingham]

Think of 36 MP as a 3x larger image size than 12MP in area (which is the case using same resolution output device). In terms of stops that means about 1.7 stop in favor or larger image, thus reducing it to 12MP, in principle, would gain 1.7 stop in all areas, in terms of perception at least.

Now, if you are thinking in terms of the D7000 sensor upsized to FF, it would be 36MP. By most tests and reviews, in high ISO it is about 1 stop under the D700 and about 1.7 stop under D3s. Using DxO mark, for example, cut points for high ISO in noise terms are:

Almost exactly what I found in my testing almost a year ago. However, when actual prints of the same size were examined, the noise became less evident (D700 vs D7000 only).

D7000: ISO 1167
D700: ISO 2303
D3s: ISO 3253

THus, resizing a 36MP file with same performance as the D7000's to 12MP would provide similar noise at 3200 ISO as the D3s.
--
Renato.
http://www.flickr.com/photos/rhlpedrosa/
OnExposure member
http://www.onexposure.net/

Good shooting and good luck
(after Ed Murrow)

--
Steve Bingham
http://www.dustylens.com
http://www.ghost-town-photography.com

 

[lock]

..on the technical aspect of reducing the resolution ?

The simple way, dropping 2 out of 3 pixels and making a 36 Mp picture a 12 Mp one,

will affect every pixel and not just the ones showing noise. So you also redcue the number of pixels you do not wish to lose. So what's the gain ? Particularly if you wish to keep the same print size ?

 

[happypoppeye]

...and gotta admit, it is a good question.

I have no idea on this one.

Amazing how rumor posts get 100s of replies, but a legitimate question... nothing.

--
http://www.wanderinground.wordpress.com
http://www.pbase.com/happypoppeye

It's almost all opinion folks, gonna have to deal with it.

Equipment: 1 Finger, an eye, a camera body and 50mm (equivalent) lens, half a brain, and a lot of money leftover to spend on using that equipment

 

[tulo]

this is a simple experiment.

take a image which you find 'noisy'. from this image cut one 100x100 pixel square and then another 173x173 pixel square. viewed at 1:1 these two patches should be 'equally' noisy.

now downsample the 173x173 square to match 100x100 using whatever algorithm you see fit and compare this to the unscaled 100x100 pixel square.

it is your perception of noise that really matters to you , no?

 

[Steve Bingham]

Please disregard what I said below. I was in error!

Let's look at it another way. If you take a 12mp picture and uprez (resample) it to a larger size you will introduce more noise - if the print size is large enough to see it! Doing the opposite should reduce the noise. However, I don't think the noise reduction will be as much as you might expect. Having no files from a 36mp camera available, it is difficult for me to test. And . . . I always prefer testing verses theory.

Below requires some explanation. If we are talking two different sensors, then yes noise would obviously be less. If you are talking about resampling, the noise would be the same.

As all pixels are not created equal this then becomes non answerable. However, if the pixels have identical qualities then yes, noise will be reduced.

Can you be more specific as to which sensors you are talking about? There are people on this forum that know far more about sensors than I do!
--
Steve Bingham
http://www.dustylens.com
http://www.ghost-town-photography.com

--
Steve Bingham
http://www.dustylens.com
http://www.ghost-town-photography.com

 

[Kaj E]

Assume you have two cameras with the same sized sensors and technology as well as the same relative pixel design, one 12 Mp the other 36MP.

In this case the noise for the same size prints would be similar. You could consider the 36MP image downsized to 12 MP as "binned" in camera. The only difference would be that the higher noise per pixel in the 36MP camera would disturb the downsizing process and in this way induce slightly more noise.

So the short answer is that noise mostly depends on sensor size technology being equal, not so much on pixel density.

Hi,

If have two images, one 12MP and one 36MP and they both appear to have the same noise when viewed at 100%, how much noise will I eliminate if I down-size the 36MP image to 12MP and view it at 100%? If I now put both these 12MP images side by side, will the down-sized image look much more clean?

Thanks!

--
Kind regards
Kaj
http://www.pbase.com/kaj_e
WSSA member

It's about time we started to take photography seriously and treat it as a hobby.- Elliott Erwitt

 

[cooldragon]

Agreed with your last sentence.

And assuming the sesor technology has been improving, 36MP images down-sized to 12MP images should give better results if the sensor sizes are the same.