Canon 10D is blessed with a reasonably low noise level, but still it becomes somewhat noticeable at ISO 800. Here is a way to reduce the noise at the exposure time by INCREASING the ISO.

Looking through the Phil's DPreview of 10D http://www.dpreview.com/reviews/canoneos10d/page18.asp I noticed that the noise is increasing slower than the ISO. In particular, when the ISO changes by a factor of 2 from 800 to 1600 the noise level increases only from 2 to 2.5, that is only by a factor of 1.25!

Based on that observation, I thought that instead of taking a picture at 800, I can increase the ISO to 1600 and dial +1 stop compensation to overexpose the sensor by 1 stop, that is by a factor of 2. (Don't dial the compensation when you are in Manual, of course, just keep the settings the same). Then (I presume that shooting is in RAW) one can dial down the exposure by one stop during the RAW development. The expected signal to noise gain would be 2/1.25=1.6 compared to just shooting at 800.

Below is an example that illustrates that. A grey card was shot with a 50/1.4 lens, indoors, RAW, AWB, focused at infinity. The first patch was shot at ISO 800 and 1/200s f/4.0, which was the normal exposure for the ambient light. The middle patch was at ISO 1600 and 1/200s f/4.0, that is overexposed by 1 stop and then brought down by 1 stop during development with C1. The right patch was shot at ISO 1600 and 1/400s f/4.0, that is normally exposed.

To my surprise, the middle patch exhibits the lowest amount of noise as visually, as well as measured by Noise Ninja.

This makes little to no sense to me... How does the +1 help ? Is the noise more visible in the darker colors, so that is why this is helping ??

Yes, you might say that, the noise is more noticeble at lower light levels. If you just shoot at 1600, normally exposed, then the noise increase is huge - see the right patch. By overexposing at 1600 and compensating this overexposure during RAW development, the noise goes down - see the middle patch.

Interesting observation! Have you tried camparing ISO 1600 to 3200 pulled 1 stop? I wonder how much worse than 800 it would be for the extra stop...

You may be on to something.... I just took two shots (RAW ) at ISO 1600, one was pushed by 1 stop and the other was not.

Brought them both into Photo shop CS and did nothing but pull the overexposed one
down by -1.

Then zoomed into a dark area to see the noise. There was much less noise in
the pushed picture. and even more detail shows in it also...

pushed....

Jack Flesher wrote:
Interesting observation! Have you tried camparing ISO 1600 to 3200 pulled 1 stop? I wonder how much worse than 800 it would be for the extra stop...

Haven't done it yet. Also, as I recall, 3200 on 10D is not a real ISO, it's actully 1600 at 1 stop compenstaion so, I'm afraid it will be a wash.

You can overexpose a grey card, but not real life...

That's whats making the difference here.

The bottom line is to use the lowest ISO you can,
while getting the most light possible to the sensor.

skid00skid00 wrote:
You can overexpose a grey card, but not real life...

That's whats making the difference here.

The bottom line is to use the lowest ISO you can,
while getting the most light possible to the sensor.

Totally agree with you.

Unfortunately, often there is a limit imposed on how much light you can get.

A specific example: You're shooting an indoor baskeball game and the normal exposure at ISO 800 is 1/200s f/2 . You can't increase the exposure - it's already the maximum lens opening and the shutter speed already borderlines with the motion blur. If you just dial ISO 1600, the normal exposure changes to 1/400s and f/2.0 So far so good but at the expense of increased noise level. Bring down the shutter speed back to 1/200s and keep shooting. That would overexpose the sensor by one stop but it can be compensated during RAW development. The final result - less noise that at the original settings ISO 800 is 1/200s f/2.

Hope, my intentions are clear.

Gennady

kites4 wrote:
Here is the pushed ISO1600 picture not zoomed.... the other two pictures is of his
hair braid.... remember the 1D has much more noise than the 10D to begin with
so this isn't too bad at all.... and this was shooting in a dim room, hand held.

not the best pic, but good for testing noise out here.


Wow, nice kitty!

gfiksel wrote:
Jack Flesher wrote:
Interesting observation! Have you tried camparing ISO 1600 to 3200 pulled 1 stop? I wonder how much worse than 800 it would be for the extra stop...

Haven't done it yet. Also, as I recall, 3200 on 10D is not a real ISO, it's actully 1600 at 1 stop compenstaion so, I'm afraid it will be a wash.



Yes, but the push is "in camera" while the pull is in CS Almost all digital cameras have a "native" ISO, and ISO increases are simply "pushes" from that base ISO. I am thinking it is the processing that goes on during the pull in CS that lowers the noise.

Just a thought...

Jack

I bet the noise is even less if you shoot at 2 stops 'over-exposed' and pull the image back '2 stops' during raw processing, for the excercise of shooting a grey card.

The real lesson is to shoot with an exposure setting that is as close to 'over-exposed without losing any details' as much as possible. The more 'over-exposed' the image is (still without losing details), the lower is the noise level. That is why this is a nice excercise in shooting a grey card, but hard to implement in real world when the tonal range of a scene is much higher than 1 stop.

Edited by slau on Feb 29, 2004 at 11:49 AM GMT

Isn't this the same thing as "exposing to the right" as discussed by Michael Reichmann?

See

http://luminous-landscape.com/tutorials/expose-right.shtml

Ummm, if you overexpose by a full stop, you're probably going to blowing highlights out the wazoo.

This assumes you have a scene with any decent amount of dynamic range to begin with, and that you were properly exposing to begin with.

slau wrote:
I bet the noise is even less if you shoot at 2 stops 'over-exposed' and pull the image back '2 stops' during raw processing, for the excercise of shooting a grey card.

The real lesson is to shoot with an exposure setting that is as close to 'over-exposed without any details' as much as possible but still retains all details. The more exposed the image is (still without losing details), the lower is the noise level. That is why this is a nice excercise in shooting a grey card, but hard to implement in real world when the tonal range of a scene is much higher than 1 stop.

Agree.
That's why 2 stops may be too much. Hey, even 1 stop can be to much when shooting outdoors in bright light, but then usually the noise is not a problem too
For shooting indoor sports one stop is usually tolerable.

DavidP wrote:
Ummm, if you overexpose by a full stop, you're probably going to blowing highlights out the wazoo.
.

Not necessarily, shooting in RAW makes it possible. Not always, but still..

The good news is that this process can still be useful in a LOW light situation, where (as David points out) you aren't blowing highlights -- or at least important ones fully. And I guess in this regard it is essentiually the same as Reichmanns "expose to the right" concept when relegated to extremely low light levels... Or maybe we should call it "exposing for the shadows", since in very low light a significant portion of the image is generally in a shadow .

MikeBinOKlahom wrote:
Isn't this the same thing as "exposing to the right" as discussed by Michael Reichmann?

See

http://luminous-landscape.com/tutorials/expose-right.shtml

I think so. I would even call it "exposing to over the right"

I agree with Mike B. this is the concept of exposing right. There are several articles about this topic and a very good one can be found on Roger Cavanagh's website. There is also an interview I read with the developers of Adobe Camera Raw who explain why this works. It is because half of the entire dynamic range information of a photograph comes from the right 1/5 of the historgram. Therefore, if you expose right as far as you can without blowing the highlights and then adjust EC in development you will increase dynamic range and decrease noise. I highly recommend looking up articles on this topic.

Stephen

Jack Flesher wrote:
The good news is that this process can still be useful in a LOW light situation...

Yes, and not only that but also when increasing the light throughput is unattainable or undesirable my other means (the shutter speed, the lens opening, screaming at the referees... )

This discussion can be summarized as follows: "don't underexpose when shooting at high ISO".

Gennady, the fact that you could overexpose by one stop and still keep all detail, indicates that you underexposed in your ISO 800 shot. Indeed it is good advise not to underexpose shots for a fear that using a higher ISO level would introduce more noise. Also, in dimly lit situations there is less risk of blowing out highlights.

sfowler wrote:
I agree with Mike B. this is the concept of exposing right. There are several articles about this topic and a very good one can be found on Roger Cavanagh's website. There is also an interview I read with the developers of Adobe Camera Raw who explain why this works. It is because half of the entire dynamic range information of a photograph comes from the right 1/5 of the historgram. Therefore, if you expose right as far as you can without blowing the highlights and then adjust EC in development you will increase dynamic range and decrease noise. I highly recommend looking up articles on this topic.

Stephen
Stephen,

I totally agree with what you are saying and I am very well familiar with those articles.

I am afraid of sounding like a broken record but frequently you just can't increase the available litght level. What is interesting in the given case that the concept works even when you are exposing more to the right by increasing the ISO, which is usually associated with the noise growth.

Here is a link to the Roger Cavanagh article Stephen mentioned:

http://www.rogercavanagh.com/helpinfo/28_exposeright.htm

The comments that said this could result in blown highlights are correct--So you can't use this technique in a photo which is pushing the right side of the histogram already.

This technique DOES effectively amount to increasing exposure time that you'd normally use for the ISO in question. So it may not be appropriate for action shots.

I don't have a 1-series camera yet (waiting patiently on my 1Ds from Adorama--Sigh), but seems like the audio file feature on the high-end cameras would be a good place for verbal note-taking on what you've done, including how much overexposure. Ideally, I'd be such a skilled photographer that I'd be able to deduce this from studying the RAW the image, and often I can, but I'd rather KNOW. :-)

Just a thought:
I think that there is always some sort of noise reduction in camera (it's the manufacturer know-how) and probably with higher ISO it is more aggressive, probably at the expence of something else (details likely). That could explain why the strange procedure seems to work (otherwise the S/N ratio should be invariated, since overexposing one stop and at the same time doubling the ISO setting would mean that the same amount of light reach the sensor).

This is merley a consequence of "exposing to the right." While it works to control noise, in the real world I would hope we are taking pictures of things with a bit more dynamic range than a grey card. When you have a lot of dynamic range in the picture, you probably will not be able to do this without blowing the highlights.

Interesting. I'd like to see as one of your grey card examples a shot at 800 ISO, overexposed by one stop, then brought back. Also, maybe doing such a test on a color swatch card (with grey scale) might be more informative as to actual real-world results (as far as exposure issues with highlights go).