Postprocessing of underexposure images vs correct exposure with high ISO?

[Passakorn]

Hi,

I am usually in a situation I have to take a bird photo under a dense forest with very low light. The typical correct exposure would be around f6.3, 1/250 and ISO is already around 3200 (Z8/9 with 180-600/800PF on a tripod). Most of the subjects are birds so I cannot go much lower shutter speed than that to maintain a sharp image because birds are moving around. There are times that this setup can be at least 3-5 stops underexposure.
I am wondering which would provide a better image quality between:
1. taking the photo with 3 stops underexposure and then postprocess with photoshop etc
2. Increasing the ISO for 3 stops (i.e. 25600) and then taking photos with a correct exposure?

My subject is rather small, with the height of the birds about 1/4-1/3 of height of the screen.

Thank you so much,

Passakorn

 

[DRwyoming]

Thank you so much,

With most if not all modern cameras it really doesn't make any difference. Modern sensors are ISO Invariant across most of their ISO range especially the higher portions of the ISO range. Changing by one, two or three stops in-camera is the same as making those same ISO changes in post processing.

Personally I'd try to get the exposure fairly close to correct in-camera. Among other things the auto focus system tends to work better when the exposure is close to correct but it's also easier to see the image, the composition and the like when you get close to the correct exposure in-camera rather than shooting a very dark image and correcting it in post.

 

[Cameron T]

Everything DRwyoming said I agree with.

 

[David Berry]

Which would provide a better image quality?
1. Taking the photo with 3 stops underexposure and then postprocess with photoshop etc
2. Increasing the ISO for 3 stops (i.e. 25600) and then taking photos with a correct exposure?

#2

 

[Replytoken]

Looks like both the Z8 & Z9 have a bump in DR at ISO400 so I believe that indicates a dual gain (ISO invariant) sensor. This gain is a little less than a stop, but if you happen to be near that ISO, which I believe you will almost always be over given your description of your shooting conditions, then you want to be at or above ISO 400 to take advantage of the gain.

--Ken

 

[EricBowles]

The Z8 and Z9 are largely ISO invariant, so it does not make much difference whether you underexpose moderately and adjust in post, or increase ISO when shooting. The one big thing that suggests using the proper exposure and ISO is the impact on focus. I've found that subject detection and focus in general work much better with a normal exposure, and with underexposure of birds you lose subject detection. It starts with losing eye detection, but you may also lose head detection and fall back to the old technology where the camera just finds something with contrast in or near the focus area you are using.

 

[bleirer]

You could also split the difference, with the advantage of being able to see the scene better in camera while keeping safe any highlights that might be at risk of being blown, for example a shaft of strong sunlight filtering onto a light colored subject at just the right angle. Or in general keep the histogram well away from the right wall.

Good article on invariance if you want to get in the weeds:

ISO Invariance Explained

In this article, we will explore ISO invariance in detail and how you can use it to your advantage in your photography

photographylife.com

 

[John Woodworth]

The only thing that matters is the amount of light---the number of photons--hitting your sensor. Assuming your aperture is wide open, going up in ISO will cause the camera to suggest a shorter shutter duration, and hence less light will fall on the sensor, and thus there will be less dynamic range (manifested by more noise) in the image.

The usual suggestion is to start with a proper exposure (and ISO) so that you get a sharp image with the camera focusing accurately and no motion blur. Then, gradually reduce ISO (and hence shutter speed) and shoot bursts until you can no longer freeze motion.

Also, if you are on a tripod and there is no motion in your background, you might be able to shoot a long exposure for the background, and then shoot a short, motion freezing, exposure for the subject (which hopefully is in better light) and then blend them in Photochop. I mean, what else you gonna do while you just sittin' there waitin'?

Also, ISO 3200 is not all that high given modern post processing techniques.

I agree with most of the preceding replies.

 

[bleirer]

Another approach is to shoot manual with auto iso and choose the ideal shutter speed and f number for your creative vision, then evaluate the indicated ISO. If iso is too high for you then see if there is a compromise that works in shutter speed or f number or shoot the background seperately to be merged later or simply accept that there will be lost shadows.

 

[Niels.hr.Hansen]

I would definitely go for getting the exposure as right as possible in camera.
As I have understood the in-camera processing, the signal picked up from the sensor goes through an analogue amplification witch is quite good at distinguishing between noise and signal. This means that it primarily increases the actual signal.
If you underexpose you don't get the same benefit of the analogue amplification, when increasing the exposure in post-proc the amplification will hit both noise and signal with same amount and are not improving signal-noise ratio. All-in-all it will provide an image with more significant noise elements.

With subject only taking up 1/4-1/3 of the image, it will be tempting to crop the image. Cropping will increase the level of noise in the final image. So, theere is also a question whether it could be a good idea to get an extender. Of course it will reduce lights ending up at the sensor, but it will remove a potential need for cropping.

 

[PhotoLover]

I would definitely go for getting the exposure as right as possible in camera.
As I have understood the in-camera processing, the signal picked up from the sensor goes through an analogue amplification witch is quite good at distinguishing between noise and signal. This means that it primarily increases the actual signal.
If you underexpose you don't get the same benefit of the analogue amplification, when increasing the exposure in post-proc the amplification will hit both noise and signal with same amount and are not improving signal-noise ratio. All-in-all it will provide an image with more significant noise elements.

With subject only taking up 1/4-1/3 of the image, it will be tempting to crop the image. Cropping will increase the level of noise in the final image. So, theere is also a question whether it could be a good idea to get an extender. Of course it will reduce lights ending up at the sensor, but it will remove a potential need for cropping.

I try to get it right in camera as well -- especially for things like static landscapes and such. But with action wildlife scenes, and a strong desire to avoid blowing the highlights (egrets and other birds with a lot of white ... ) I often have slightly to sometimes alas more than slightly, underexposed images.

Which is where I think the esteemed Eric Bowles is quite right -- most modern cameras are largely ISO invariant, so IF you shoot in RAW mode then raising the shadows later in post-processing will look the same as if you had a brighter exposure in-camera. I mean, you can't be vastly far off and expect much, but a stop or two usually doesn't matter. You do have to be careful about underexposure of an image with a lot of shadow area ...

 

[Archibald]

In dim situations I often shoot at crazy-slow shutter speeds in burst mode. Usually you can get a sharp shot in the set, and at a reasonable ISO.

The shot below is an example. Hand-held and heavy crop. The first one is as-is, the second one was enhanced in Lightroom.

You can only see EXIF info for this image if you are logged in.

You can only see EXIF info for this image if you are logged in.

 

[Replytoken]

As I have understood the in-camera processing, the signal picked up from the sensor goes through an analogue amplification witch is quite good at distinguishing between noise and signal. This means that it primarily increases the actual signal.
If you underexpose you don't get the same benefit of the analogue amplification, when increasing the exposure in post-proc the amplification will hit both noise and signal with same amount and are not improving signal-noise ratio. All-in-all it will provide an image with more significant noise elements.

Tis is the first I have heard this in all of the ISO invariant discussions I have read. Many times I recall folks saying that a computer software package has a lot more power and ability to adjust an image than the chip in the camera. Not really sure I find either extreme on this spectrum totally correct, but it would be good to know which approach has advantages the other does not.

--Ken

 

[bleirer]

Tis is the first I have heard this in all of the ISO invariant discussions I have read. Many times I recall folks saying that a computer software package has a lot more power and ability to adjust an image than the chip in the camera. Not really sure I find either extreme on this spectrum totally correct, but it would be good to know which approach has advantages the other does not.

--Ken

Both are true but it depends on the camera. Modern high end cameras after iso 400 or so it makes no difference where you brighten it. Older cameras it did. Check out the article I posted above.

 

[ajrmd]

Well, the answer is a bit more complicated than a simple binary yes/no. The Z8/Z9's sensor is a dual gain (64,503) and it is fairly ISO invariant across its range. Nonetheless, all RAW images have to be processed through converters which introduce their own biases (curves, sharpening, etc.) to the image. If you use linear RAW to minimize that effect then within a reasonable exposure range, the noise will be similar. For example, here is a screenshot of two images, one exposed correctly at ISO 3200 (right) and the other one shot a second later at ISO 500 (-2.67 stops if I did the math correctly) and then the exposure corrected by that amount during RAW conversion. The screenshot is small and highly compressed to fit here, but I can assure you there is no substantial difference in noise. As others have discussed, the primary reason to expose correctly with a ISO invariant MILC has more to do with being able to see the subject in the viewfinder and for the AF to work properly.

You can only see EXIF info for this image if you are logged in.