Reading mode: Light Dark

Guess what ISO...

Go to last post
photofan1986

photofan1986

Senior Member
Messages
4,123
Solutions
3
Reaction score
2,954
Location
BE
...I shot this pic!
Used my 450D, and a special trick + some PP. Resized to 3000x2000
Hint: the upper limit of the 450D is not 1600ISO in this case.

So what are your guesses?



 
First a question: The displayed image is 900 x 600 pixels, not 3000 x 2000. Is the 900 x 600 the full-frame image? If so, I'll guess it's an effective ISO 6400. Why? Here's a similar one I shot with my XSi at an effective ISO 12,800 reduced to 900 x 600 pixels.

 
Click the (original) button to see it 3000x2000 ;)
 
The upper limit of the 450D is 1600. There is absolutely nothing you can do about that.

You can underexpose an ISO1600 image, then make it brighter in post-processing and losing some bits (the bits for the highlights are not used) and then perform strong noise reduction. It's still an ISO 1600 image.

Cameras like the 550D have a setting that says ISO 6400 but does the sensor really have ISO 6400 or is it an underexposed ISO1600 or 3200 image that is brightened up? I don't know exactly what their real upper limit is.

Whatever you do to end up with the image you want is fine, but thinking that you are changing the ISO in doing so is an illusion.
photofan1986 said:
...I shot this pic!
Used my 450D, and a special trick + some PP. Resized to 3000x2000
Hint: the upper limit of the 450D is not 1600ISO in this case.

So what are your guesses?



Click to expand...
--
Slowly learning to use the 450D, the Canon G6 and the Fuji F200.
Public pictures at http://debra.zenfolio.com/ .
 
It looks great small, but when I look at the original I see a disturbing texture superimposed on the image. It is most noticeable in the wood at the right. Is this an artifact of the upload or part of the "trick" you mention?
 
Looking at the full sized image there is a definite 'granular' look to the image?

At the smaller size it is not so obvious.

--
Phil
 
Yours is way better than my ISO-effective-12,800 at 3000x2000. I'm eagerly awaiting to hear your effective ISO and how you did it.
photofan1986 said:
Click the (original) button to see it 3000x2000 ;)
Click to expand...
 
Paul De Bra said:
The upper limit of the 450D is 1600. There is absolutely nothing you can do about that.

You can underexpose an ISO1600 image, then make it brighter in post-processing and losing some bits (the bits for the highlights are not used) and then perform strong noise reduction. It's still an ISO 1600 image.

Cameras like the 550D have a setting that says ISO 6400 but does the sensor really have ISO 6400 or is it an underexposed ISO1600 or 3200 image that is brightened up? I don't know exactly what their real upper limit is.

Whatever you do to end up with the image you want is fine, but thinking that you are changing the ISO in doing so is an illusion.
Click to expand...
Well, yes, the upper limit of the 450D is 1600ISO, but this is only a software limit. It's basically the same on a 500 or 550D. The only difference is that the on a 550D, you can select those values in the menu, whereas on a 450D, you have to SHOOT RAW, underexpose and then COMPENSATE THE EXPOSURE (and not the brightness).

Basically, the sensor only has ONE sensitivity: the basic one. On a 450D/500D/550D, this value is 100ISO. On some Nikons, it's 200. What the processor in the camera does is basically the same. It amplifies the signal coming from the sensor when a higher ISO is selected. You can emulate this by underexposing and correcting the EV in RAW.
 
OK, well this picture was shot at at corresponding ISO of +-15.000.

It was shot at 1600ISO, underexposed by 3 stops, and later I saw that the picture still was too dark, so I had to add half a stop, which bring the actual ISO to +- 15.000.

Now, how did I manage to get such a good result? :D
 
photofan1986 said:
OK, well this picture was shot at at corresponding ISO of +-15.000.

It was shot at 1600ISO, underexposed by 3 stops, and later I saw that the picture still was too dark, so I had to add half a stop, which bring the actual ISO to +- 15.000.

Now, how did I manage to get such a good result? :D
Click to expand...
You did get a good result there, well done.

The very fine grain nature of the noise (when viewed at 100%) reminds me a lot of the new Adobe Camera Raw when used within Lightroom3.
 
I'm not sure it shows anything really.

High ISO noise only happens in conditions for which the processor really needs to amplify the incoming signal, as you legitimately pointed out. So a well lit scene will in practice show little noise, because the camera reduces Tv (or the aperture) in order to compensate for the excess light hitting the sensor.

Bottom line, real high ISO capability is tested only for a combination of high (exposure time) X (aperture)

The 450 D does NOT have more than 1600 usable ISO (and actually probably fails well before)... Unless you accept that high ISO crippled with noise is usable.
photofan1986 said:
...I shot this pic!
Used my 450D, and a special trick + some PP. Resized to 3000x2000
Hint: the upper limit of the 450D is not 1600ISO in this case.

So what are your guesses?



Click to expand...
 
Try to make a similar picture at 15.000 IS equivalent and then you talk.
 
photofan1986 said:
BlueHotel wrote:

The 450 D does NOT have more than 1600 usable ISO (and actually probably fails > well before)... Unless you accept that high ISO crippled with noise is usable.
Click to expand...
That's BS. ISO is limited on 1600 in the camera menu, but the only real ISO is the one that allows you to shoot with a higher shutter speed. If I have a shutter speed of 1/15 at 1600ISO and then I underexpose the shot by one stop, I get a 1/30 shutter speed. Then I have to correct the exposure by +1 stop in the raw development, which is equivalent to having 3200 ISO. And so on.

Here is a picture I took at 1600ISO with my crappy kit lens 18-55IS at a very slow shutter speed. Still, it is quite sharp but I had to correct the exposure by 0.5 stop in ACR. The result below (resized to 3000x2000 for convenience).

Is that what you call a "high ISO crippled with noise"? If so, you never shot with a film faster than 100ISO, because even at 100ISO, you would get more grain than this.



 
Sensors do not have one and only base ISO. They have embedded amplifiers in each photosite and by setting them to an appropriate gain level you change the ISO of the sensor.

Some camera's like 500D offer ISO expansion. This is the case where to processor has to post process the image to expose properly.

For example when you set ISO from 100 to 3200 in 500D the processor will always get a correctly exposed image (assuming you set shutter and aperture correctly). When you set ISO to 6400 or 12800 though, the processor will get an underexposed image and will have to lighten it up.

Nevertheless your images really do look great. Congratulations! I am really curious about the technique you used. Did it involve frame stacking? I've seen a software where you shoot many high ISO images of the same subject and the program merges them to a low ISO image.

--
DO NOT PANIC!
http://www.andmarios.com
My 5 cents: turn off Highlight Tone Priority!
 
Thanks for the thorough explanation!

You nailed it: image stacking. I shot 8 raw pictures at 1600 and underexposed them by a bit more than 3 stops, using camera raw 6.1. Then, I stacked them using Traumflieger (originally, a DRi software). And finally, I used Topaz Denoise 5 to reduce the remaining noise.

Of course, I cheated. But the interesting thing to note is that image stacking allows to reduce noise AND increase detail (because noise is random) !
 
BlueHotel said:
I'm not sure it shows anything really.

High ISO noise only happens in conditions for which the processor really needs to amplify the incoming signal, as you legitimately pointed out. So a well lit scene will in practice show little noise, because the camera reduces Tv (or the aperture) in order to compensate for the excess light hitting the sensor.

Bottom line, real high ISO capability is tested only for a combination of high (exposure time) X (aperture)
Click to expand...
Exposure time (or amount of light) has almost no influence on noise levels. In short, quantity of light at a given ISO and visual exposure isn't too relevant, quality (temperature) more relevant.

There isn't more amplification when the shutterspeed is longer. The longer shutterspeed already compensates the amount of light hitting the sensor. For 2 scenes that both equally fill the histogram, the first with an exposure of say 20s F4, ISO1600 and the second 1/200s F4 ISO 1600, will thus show similar levels of noise. Because the light hitting the sensor was similar too.
 
TrojMacReady said:
Exposure time (or amount of light) has almost no influence on noise levels. In short, quantity of light at a given ISO and visual exposure isn't too relevant, quality (temperature) more relevant.
Click to expand...
This is so very wrong. Haven't you read the many "why are my images so noisey?" posts on this site? And the answer is nearly always "they are under exposed". And have you ever heard of "expose to the right"? That is expose your image so your histogram (when shooting RAW) is as far to the right side without clipping the highlights. This will lead to less noisey shadows because you are exposing the shadows as well as possible. You can then pull down the exposure during RAW development to balance out the contrast.
TrojMacReady said:
There isn't more amplification when the shutterspeed is longer. The longer shutterspeed already compensates the amount of light hitting the sensor. For 2 scenes that both equally fill the histogram, the first with an exposure of say 20s F4, ISO1600 and the second 1/200s F4 ISO 1600, will thus show similar levels of noise. Because the light hitting the sensor was similar too.
Click to expand...
No but under exposure leads to more noise, which is why very high ISO gets noisier because the camera is essentially under exposing "more" and pushing the exposure further!
 
RedFox88 said:
TrojMacReady said:
Exposure time (or amount of light) has almost no influence on noise levels. In short, quantity of light at a given ISO and visual exposure isn't too relevant, quality (temperature) more relevant.
Click to expand...
This is so very wrong. Haven't you read the many "why are my images so noisey?" posts on this site? And the answer is nearly always "they are under exposed". And have you ever heard of "expose to the right"? That is expose your image so your histogram (when shooting RAW) is as far to the right side without clipping the highlights. This will lead to less noisey shadows because you are exposing the shadows as well as possible. You can then pull down the exposure during RAW development to balance out the contrast.
Click to expand...
I clearly stated "given (equal) visual exposure" and below "equally filling the histogram" .

If the light is equally distributed over the histogram for both physically different exposures (not to confuse with visual exposure), then there shouldn't be any or hardly any difference in noise. This has been tested and demonstrated ad nauseam. It's simple physics really. The same amount of photons hits the sensor, that's why cameras can change aperture and shutterspeed to begin with....
RedFox88 said:
RedFox88 said:
There isn't more amplification when the shutterspeed is longer. The longer shutterspeed already compensates the amount of light hitting the sensor. For 2 scenes that both equally fill the histogram, the first with an exposure of say 20s F4, ISO1600 and the second 1/200s F4 ISO 1600, will thus show similar levels of noise. Because the light hitting the sensor was similar too.
Click to expand...
No but under exposure leads to more noise, which is why very high ISO gets noisier because the camera is essentially under exposing "more" and pushing the exposure further!
Click to expand...
And where did underexposure come into play? Read again.
 
Actually temperature is only relevant for long exposures and comes from so called dark current which is affected by temperature.
At shorter exposers there are 2 components to noise.

Read noise, which is noise added as the pixels are read out and which does not depend on the magnitude of the signal being read. This shows up in dark areas where there is less signal and signal:noise is lower. Read noise tends to decrease with increasing gain in modern cameras.

Photo noise is the inherent randomness in the amount of photons hitting the pixel and is equal the square root of the number of photons. This noise increase with ISO. Say you have a stream of 10000 photons at a gain of 1, that gets you an output luminance of 10000 with noise of 100, so signal to noise of 100. Instead increase the gain to 25, to get the same output (brightness in final picture) you would reduce the exposure by 25 ie 400 photons, which would give you 20 photons of noise. But that is also multiplied by the gain giving you 500 output noise. Signal to noise of 10000/500 = 20.

http://www.clarkvision.com/articles/index.html is a great source of info on the physics of photography.

Edward
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top