Dual ISO vs Auto ETTR

Started by RenatoPhoto, August 22, 2013, 08:20:56 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

RenatoPhoto

By increasing the highlight ignore in the Auto ETTR module it is possible to significantly improve the quality of the shadows in situations where the dynamic range of the scene is beyond the capabilities of the camera sensor.  The cost is the lost details in the highlights.

In my testing, I found that the DUAL ISO module increased the dynamic range of the camera but the noise in the shadows and sharpness in the brights is better by using Auto ETTR technique if some of the highlight details can be sacrificed.

If you must capture all highlight details and are not bothered by noise in the shadows and some loss of detail in the brights then DUAL ISO is capable of delivering the goods.
Here is the setup I used to test Dual ISO vs Auto ETTR


With Canon metering (Evaluative metering) the image has very deep shadows with lots of noise and the highlights were slightly overexposed



With the Auto ETTR (AETTR) module enabled in default settings (02.% h.i.) there is a slight improvement in the highlights (better exposed) while the shadows are a bit deeper.  This has been my experience with AETTR, if there are highlights, then it always meters them accurately to prevent any highlight overexposure.  I all cases where there is an excessive Dynamic Range then the AETTR will always underexpose the image to protect the highlights.  To counteract this issue, if the highlights are not so important, then I raise the highlight ignore setting until the image looks right.  In most situations with bright clouds I can get away with a 5 to 10% in the highlight ignore and still get good results.

AETTR default (0.2% h.i.)



AETTR  +5%



AETTR +10%




Now if we use the DUAL ISO module we will see how the shadows are lifted while the highlights are not overexposed and the benefits of increasing the second ISO.

Dual ISO 100-200



Dual ISO 100-400  Notice that the shadows are lifted slightly more than 100-200 image



Dual ISO 100-800  Notice that the shadows are lifted slightly more than 100-400 image



ADVANTAGES AND DISADVANTAGES DUAL ISO

In the following comparisons I have brought all of the RAW photos into Photoshop via ACR 7.1  and used the Auto exposure function which automatically does its magic to raise the shadows and lower the highlights to some predetermined "magical value".  I use this to quickly bring all the images to the same magical look as a standard point for comparison.  In am hopefully comparing images that have been adjusted to the same standard look of "Auto".

The following comparisons show various areas of the images where each feature will show its advantages and disadvantages.  The side-by-side comparisons will be made in the highlights, brights, Mids, and Low lights areas of the photo.  I have not used any technical definition for these exposure areas just my perception of what is highlights, brights, Mids, and Low lights.

Here are the camera settings for each photo:

DUAL ISO 100-200,  1/25,  f/14
DUAL ISO 100-400,  1/25,  f/14
AETTR +5%,  1/25,  f/14, ISO 100
AETTR +10%,  1/25,  f/13, ISO 400

Note: I did not record the exposure compensation for each image.

First, look at the highlights:


By increasing the highlight ignore of the AETTR module obtained the following comparison between Dual ISO and  AETTR.



Clearly at AETTR+10% we lose some of the details in the fog, in this case this does not represent any significant loss in my perception for this image.  AETTR+5% looks better but not as good as ISO 100-200.

The Dual ISO feature has conserved the highlights very well.

Second, lets consider the brights:


A problem of DUAL ISO  here becomes evident (due to aliasing) with the loss of resolution in the high contrasting areas of the chart.  This is probably due to the aliasing reduction algorithms which reduce the details at the expense of resolution.   The sharpness defects become much more pronounced with the 100-1600 and 100-3200.


Third, now looking at the Mids and low lights:

In this area of the image the noise becomes more evident and the clear winner is the AETTR+10% (at ISO 400).  The Dual ISO images are not so good and there is some horizontal noise banding appearing in the smooth brick areas.  Look below the "1" on the 100-400 image.  This banding becomes much more apparent after denoising and sharpening.



Fourth, finally the Low Light area:

The winner is AETTR+10% with less noise. Horizontal banding is still evident in the Dual ISO images.



I have also compared the AETTR+10% with the 100-800 and the 100-1600 and the 100-3200.  In the image below I raised the shadows by 10% to get a bit more detail.  Again the AETTR+10% is a very good results giving higher sharpness and very little noise.  The 100-3200 is also very good but I would not consider this option as ideal since it has lost too much details in the brights due to aliasing.  Note that horizontal banding is not evident at these higher ISOs.



MY FINAL CONCLUSIONS


1. If you have a scene with large dynamic range and need the details of the highlights and shadows then Dual ISO will be the only option.
2. If you can sacrifice a bit of details in the highlights then AETTR is better suited for the job since the noise is lower and there are no problems with aliasing.

I hope that my analysis is correct and if there is something wrong I would like to know what I did wrong to correct it!





http://www.pululahuahostal.com  |  EF 300 f/4, EF 100-400 L, EF 180 L, EF-S 10-22, Samyang 14mm, Sigma 28mm EX DG, Sigma 8mm 1:3.5 EX DG, EF 50mm 1:1.8 II, EF 1.4X II, Kenko C-AF 2X

Audionut

Another excellent post RenatoPhoto  :)

Try some green channel clipping also.

If you're highlights aren't overly important and you have a large EV difference between the highlights and important midtones, use dual ISOs further apart.
Raw zebras have been updated.  Where the overexposure zebras are solid, both ISOs have been over exposed.  Where zebras are weak, only the higher ISO has been blown.  For maximum dynamic range while keeping maximum resolution in important detail, use higher recovery ISO until zebras appear on that detail.

RenatoPhoto

Quote from: Audionut on August 22, 2013, 08:48:49 PM

Try some green channel clipping also.

If you're highlights aren't overly important and you have a large EV difference between the highlights and important midtones, use dual ISOs further apart.

Raw zebras have been updated.  Where the overexposure zebras are solid, both ISOs have been over exposed.  Where zebras are weak, only the higher ISO has been blown.  For maximum dynamic range while keeping maximum resolution in important detail, use higher recovery ISO until zebras appear on that detail.

Wow!!!  Thanks for the tips, there is so much to learn that I have to challenge myself to do things like this really understand how to use the Magic Tools!!

I will have to dig further on your suggestions!

Thanks
http://www.pululahuahostal.com  |  EF 300 f/4, EF 100-400 L, EF 180 L, EF-S 10-22, Samyang 14mm, Sigma 28mm EX DG, Sigma 8mm 1:3.5 EX DG, EF 50mm 1:1.8 II, EF 1.4X II, Kenko C-AF 2X

Africashot

Thank you so much for elaborating! It saved me tons of trial and error tests of my own, can't wait to fully embrace this and say goodbye to time consuming HDR algorithms forever!
ML 5D2 & T3i

a1ex

My first reaction after reading this: why not use both?

I'd like to see a test that confirms (or infirms) the dynamic range gained. For example, if ML says you'll get 3 EV of extra dynamic range (let's say at ISO 100/1600), take another picture at ISO 100 with a shutter speed slowed down by 3 EV, and compare the shadows. If the theory is right, the noise levels should be comparable.

RenatoPhoto

Quote from: a1ex on August 25, 2013, 11:47:56 AM
My first reaction after reading this: why not use both?

Do you mean both Dual ISO + AETTR?

Quote from: a1ex on August 25, 2013, 11:47:56 AM
I'd like to see a test that confirms (or infirms) the dynamic range gained.

Do you mean DR gained by DUAL ISO?
http://www.pululahuahostal.com  |  EF 300 f/4, EF 100-400 L, EF 180 L, EF-S 10-22, Samyang 14mm, Sigma 28mm EX DG, Sigma 8mm 1:3.5 EX DG, EF 50mm 1:1.8 II, EF 1.4X II, Kenko C-AF 2X

1%

I've used both and dual iso with auto expo. Sometimes the iso range doesn't end up far apart.

Audionut

I like auto expo with dual ISO.  You get a correct exposure (often with highlight headroom for the dual ISO), and then the added shadow recovery.

Sticking to ISO 400 for the recovery, your getting nearly 2 stops of dynamic range there, without the pitfalls of having an extended range between the 2 ISOs.

Karmaschinken

Hello, thanks a lot for you work!

There is one thing I don´t understand: I have been out to test ETTR, I was doing some landscape shots. Now theoretically the brightest areas in a scene should be those that make ETTR calculate the exposure settings, right?

If so, why is it then that ETTR comes to different settings when pointing the camera to a more darker point while the speculars remain in the scene?

I was pointing at the bright clouds, let ETTR do its calculation, then pointing more at the bottom, while still having the clouds in the picture. As far as I understood, ETTR should have led to same settings, but it didn´t. I thought it might have to do with the metering mode, but obviously it doesn´t, no?

Please help me get rid of my confusion! Thanks.


Audionut

Was there a change in the white level point?  Or did the exposure settings change?

If the exposure settings changed, I would say your lens vignettes.

RenatoPhoto

Quote from: Karmaschinken on August 31, 2013, 04:55:32 PM
If so, why is it then that ETTR comes to different settings when pointing the camera to a more darker point while the speculars remain in the scene?
You should present the settings that you used in ETTR.
Exposure target:
Higlight ignore:
Clipping mode:
Midtone SNR limt:
Shadows SNR limit:
Slowest shutter
link to Canon shutter

Also the comparing images could be helpful
http://www.pululahuahostal.com  |  EF 300 f/4, EF 100-400 L, EF 180 L, EF-S 10-22, Samyang 14mm, Sigma 28mm EX DG, Sigma 8mm 1:3.5 EX DG, EF 50mm 1:1.8 II, EF 1.4X II, Kenko C-AF 2X

Karmaschinken

Quote from: Audionut on August 31, 2013, 05:26:29 PM
Was there a change in the white level point?  Or did the exposure settings change?

If the exposure settings changed, I would say your lens vignettes.

Hmmm, tried again and I cannot reproduce it, strange. I must have done something wrong maybe or maybe you are right, it might really have been the vignetting. However, now the results are absolutely reliable. Sorry for the false negative, nevertheless thanks a lot for your help!

Canon eos m

RP: How do I blow up the samples you have in your first post above. Wanted to get a feel of the comparison by replicating the test on  my own.

My experience till now has been that ETTR though more reliable of the highlights does take a bit more learning to master. It also takes longer for the camera to meter the scene since it has to figure out the right light input for each shot. The most frustrating bit about the camera under ETTR control is that it behaves like it is under P mode. You can control the minimum shutter but do not have the same option for the aperture and ISO.

Plus, the fact that you have to pre-judge an image to decide whether it will be okay to sacrifice the shadows a bit or not is a difficult one to make. I somehow prefer the Dbl click option for ETTR.

Plus, since I mostly use my camera for portrait shots, using ETTR is risky. I find it extremely difficult in high DR situations to get rid of the black trails near the chin and below the neck, etc.

With Dual_ISO things are must simpler. Just get the ISO combo right and you are done. Anything above 100-800 is a bit risky though.

My view is during daytime and good light situations always go for ETTR for the rest there is Dual_ISO.
Canon 5D Mark III, Gopro Hero Blacks with 3D Casing, A Few Lenses, Adobe CC 2014, MacBook Pro, Windows 8 PC, Lots of Video Rig!

Started Nuke. Loved it but then the 15 day trial ran out. Back to After Effects and loving it :-)