suggestion: automatic ETTR (Expose to the right) handling

[nyukuri]

Many users use the "expose to the right" (ETTR) technique for better N/S ratio and generally better image quality.
RAW images are lightly overexposed (+0.7 to +1.7 EV) but without clipping the highlights. In the raw converter this overexposure is then corrected with the "exposure" tool and, if necessary, with highlight recovery. The histogram is then finally spread out to fill the whole possible tonal range. The result are much better images than if the photo was exposed "correctly" in the camera.

Until today, no manufacturer has implemented a simple automatic procedure, applying exactly these steps!

THIS IS WHERE ASP COULD STEP IN AND BE THE FIRST SOFTWARE SOLUTION FOR AUTOMATICALLY HANDLING "ETTR"!!!

1. an overexposed image is imported
2. ASP reads the EXIF information (exposure bias +1 EV) and corrects it AUTOMATICALLY with the "exposure tool" (decrease exposure to -1.00 EV)
3. the histogram is then spread out AUTOMATICALLY so that the image data takes advantage of the full available range (similar to AutoLevels), just avoiding clipping

DONE!

This could be implemented as a single checkbox "ETTR" in the "exposure" and "basic adjustments" tools.

a) if the checkbox is left UNCHECKED the image is not altered by the described procedure and left overexposed, assuming that this is what the photographer intended
b) BUT IF the checkbox "ETTR" (Expose to the right) is ENABLED, ASP would follow and apply the steps described above!!!


If ASP does not implement this, let's hope a clever plugin programmer will see the potential...

[ogrizzo]

Cannot you just create such a preset combining exposure bias with appropriate auto level settings?

Ottavio

[nyukuri]

I could create Presets for automatic exposure correction, but for each EV Bias I would need another preset (+0.3, +0.7, +1.0 etc.) And since ASP cannot filter for EV Bias it is almost useless having such presets, because I need to look at the META data for each image independently and then choose the right preset - An ETTR plugin would work correctly no matter what the EV Bias is!

And thats only the first step. In a second step the histogram has to be adjusted by moving the white point. Now for this a preset is absolutely useless because the adjustment needs to be different for EVERY image, even if they all have the same EV Bias!
The AutoLevel function somehow does not work right after the first step! Can't say why, maybe it's a bug. It does cause some change but the histogram does not show that the right gap for the highlights is filled up...!

Only a program/plugin could read the histogram and move the white point down to adjust the histogram so that clipping is just avoided. If you have overexposed images, try to do these steps by hand (1. correcting exposure, 2. adjust the histogram, by hand or AutoLevel) and you will see that each image needs a different handling in the 2. step and in the 1. step as well if the images are differently exposed.

So I am sorry to say so, but presets simply cannot do with ETTR!

[nyukuri]

Ok, your post made me try again the AutoLevels plugin, and with that I could do the second step, because it is placed later in the pipeline. The ASP AutoLevels comes too early to be useful in this case. So thank you making me aware of this.
What remains is now only the first step, the auto exposure correction. But here remains the problem that the Metadata browser cannot filter for the EV Bias...

[grubernd]

i have some practical questions about this, because i either dont have the time to do this or expose differently anyway.

from my understanding of ettr it means that you know exactly at what overexposure the data will blow in your raw-converter, then go back about a 1/3 or 1/2 stop, expose. for that to work you need to spotmeter the brightest parts in your image, because any automatic exposure i used is either very general (mid-centered) or driven by complex algorithms (matrix), where neither deliver predictable results for small parts of the image.
now if you spotmeter the whites and go from there you should always be using the exact same exposure compensation, right? wouldnt that allow for a preset to get everything to a very predictable look?

[nyukuri]

For me only results count, and I get good results even without a light meter. I simply overexpose, usually 1/3, 1/2 or even 2/3 EV, watching the histogram on the camera to make sure highlights are not blown.
After correction in ASP (exposure correction, moving white point to "fill" the right gap of the histogram) I see significantly less noise, more details and better tonal range in shadows.

Now why do I want a plugin for this correction.
- as I said, I use different exposures. ASP cannot (yet) filter for EV Bias, so I have a folder mixed with images of different overexposure. Even using presets, I have to view each image individually and mark them in such a way that later I can apply a preset, lets say on all images with +1/2 EV.
A plugin could read the EXIF and correct exposure automatically
- The White point needs correction every time differently. ASP AutoLevels doesn't help at all, its too early in the pipeline. The AutoLevels plugin does in many cases a good job, but sometimes it simply seems not to effect the histogram as it should, and I need to work manually again.

Someone suggested to me that the newer sensors make ETTR nowadays superfluous, but I am not yet convinced. I hope to get a new camera next month and of course will test it thouroughly, really interested if my ETTR habit will no longer be necessary.

[grubernd]

ok. maybe this behaviour can be coded into CA-Quest?
i guess all the mechanisms for it should be in place there to do automatic exposure re-compensations.

about "new" cameras.. the Nikon D200 as of 2005 was one of - if not - the first camera to have pre-cooked raw-files that used a highly improved A/D conversion stage to get much better distribution of data inside the raw-files.

bibble/aftershot can show you the near-real raw data: switch the Color Mgmt setting to "None - Linear". the brighter and more similar your image is to the "Icc profiled" setting, the better the raw data and the less is gained by using the hardcore ETTR mechanisms.

[afx]

But if you switch to linear mode, remember there is currently a bug that renders the image in linear mode 2 stops underexposed...

cheers
afx

[nyukuri]

ok. maybe this behaviour can be coded into CA-Quest?

That makes sens, since CAQuest is already handling other EXIF data and applying corrections accordingly. I contacted already the author...

the brighter and more similar your image is to the "Icc profiled" setting, the better the raw data and the less is gained by using the hardcore ETTR mechanisms

well, after comparing, I now understand why I have such a difference with ETTR looks like a better sensor (camera) really is needed

[grubernd]

well, after comparing, I now understand why I have such a difference with ETTR looks like a better sensor (camera) really is needed

mind the bug that afx posted above!
i had totally forgotten about it, so currently the linear mode seems to be broken and is not really an indicator for the raw data..

[nyukuri]

sorry for not mentioning it before, but yes, I did pay attention to that bug! still there is a significant difference visible

[andrewj]

Hi,

Thanks for the plugs!

Yes, it should be pretty easy to add the auto-correction to CAQuest. I will need some sample data to check it's working. Could you please upload a couple of sample pics (which require different correction), plus the XMP files for them manually corrected as you expect. I'm just aware of some subtleties in how ASP sequences changes which might be a bit tricky.

What I'm personally interested in is trying to get the camera to do ETTR automatically. I recently bought a Panasonic GH2 and had hopes that the firmware could be hacked to provide this, but so far I haven't got anywhere, as the hacking community seems entirely focused in video. Oh well...

Let me have details of the sample files, and I'll get cracking.

Andrew

[andrewj]

Hi Guys,

I'm just looking at this. Two questions so far:
1. What do you mean by "The AutoLevels plugin"? I don't seem to have either a plugin or a separate UI tool for that. The "Exposure" tool just sets the same values as autolevel on "basics".
2. How can we distinguish between a shot which has deliberately been overexposed for ETTR, one which has been overexposed for bracketing purposes (e.g. part of an HDR triple), and one which has had positive compensation applied which we want to keep (e.g. a wedding dress)? I assume we want to apply the correction in the first case, but not the other two. Is there any other indicator?

Any ideas?
Thanks
Andrew

[nyukuri]

Thank you Andrew, some RAW files are on the way, but it will take it's time, due to my slow internet connection... may take hours, if not a day, to send just a few...

[nyukuri]

1. There is a third party plugin "AutoLevels" that comes later in the pipeline than the original ASP AutoLevels function of the Exposure Tab. You can check it on ASP's official plugin site. This AutoLevels plugin usually does work with my ETTR handling, where as the ASP AutoLevels doesn't at all.

2. Except that for bracketing there is already another EXIF info, I am not aware of how to distinguish overexposed images. I think we have to leave this up to the photographer. I don't think that someone will mix them a lot. There are rather HDR guys who do all with bracketing - they will simply not Enable your plugin for ETTR. Others, like me, use ETTR extensively, and by far not so often bracketing. So usually the photographer should know already when importing his images if he wants the ETTR handling enabled or not. I don't see there a big problem.

All you have to do is to give each correction function it's own "enabled" checkbox, not only one for the whole plugin. This way I can have the plugin correcting my lens, but NOT exposure or vice versa etc...