Wednesday, March 1, 2017

How Much Exposure Error Can Be Made With The Olympus E-M1 Mark II And The Fuji X-T2 And Still Have A Usable Image? Part II of III

Scene exposed and rendered as closely as possible as I viewed it. (click to enlarge)
Olympus E-M1 Mark II, 12-40mm f/2.8 lens @ 25mm; 1/400th sec.@ f/8; ISO 200
Apology up front:  The focus is slightly off in this series of images and I can't
explain why. I focused manually and then made the images.  My only
reasoning is that I bumped the focus ring on the lens, which was in manually coupled mode.
This is Part II of III in my testing exposure latitude and exposure error of my Fuji X-T2 and Olympus E-M1 Mark II.  This part will deal with the Olympus camera as part I dealt with the Fuji camera.  You can find Part I XXXX.

A few weeks ago I conducted some tests to determine how my Fuji X-T2 and my Olympus E-M1 Mark II differ in rendering color as I had wondered if they could be matched for color or if their color palates were so different that the two cameras couldn't be matched. That led me to wonder how Fuji Acros and Olympus Monochrome differed in the tonal renditions of the various colors.  Those being completed, I wanted to conduct one more test to better understand the sensor latitude of both cameras and how much exposure error could be tolerated and still get something usable.  "Usable" and "decent" are very subjective, but I wanted to know the limits of exposure with these two cameras.  I had conducted this particular test on before on my Nikons as well as my original Olympus E-M1.

You can read the color matching test here, the Acros versus Monochrome test here and the results of the Nikon exposure latitude test here and finally the Olympus E-M1 latitude test here.


This is Part II of a three part posting.  Part I was the testing I conducted with my Fuji X-T2.  This post, Part II will reveal the testing results with the Olympus E-M1 Mark II and Part III will be a summary and comments about these tests and my findings

What is the purpose of this particular test?  Again, I felt it was important for me to fully understand the ability of
 my cameras' sensors to handle severe underexposure as well as severe overexposure, scenes with extreme tonal ranges and still achieve a decent or usable (the operative words here) image after editing.

Disclaimer:  This is not a scientific test and I don't portray it to be one.  This is a practical assessment based upon visual interpretation and personal judgment.  There was no scientific testing equipment utilized to fully measure contrast, densities or color hues.  I standardized a testing procedure and applied it equally to both cameras to be able to compare each to one another.  In other words, your results may vary if you conducted the same test or your judgment as to what a usable or decent image is may differ from mine.

Several days ago I found the scene depicted in the image at the top of this post. I thought it would make a pretty good subject for testing exposure latitude since it had deep shadows, bright highlights with detail in the bright metal roof piece, a variety of colors, texture and subject detail.  Although it is full winter sun and the sun was at my back. The sky was clear.

I set the cameras on a tripod in sequence, set base ISO of 200 on both cameras, determined a base exposure according to the histogram, then varied the exposures up and down by six stops.  The file format was RAW and all other camera controls were at default.

When I imported the images into Lightroom, I changed the profiles (Calibration Menu in the Develop Module) of both cameras to the ones I generated using the X-Rite Colorchecker Passport. I did this so the two different cameras would have consistency in color and contrast.  I didn't want to use the Fuji film simulations or the (currently only) available Adobe Standard-Beta profile setting as those would render differently with no consistency and I felt that would hinder my ability to compare the two cameras to each other.  The Colorchecker profiles provided consistent images across both cameras.

The images below depict the results and the captions detail what I found.  Your evaluations may differ from mine, but I was looking to find the extremes in pushing the sensors as far as possible and finding the point where there was no way to bring the image back to approximately what I saw in the normal exposure.

The image at the top of this post is what I judged to be how the scene looked to me while standing in front of it, as much as it can represent reality.  We know that is not totally possible as the human eye and brain's interpretation is much different than the literal and/or electronic interpretation of a digital sensor and associated processing algorithms.


(One thing I discovered during the creation of the below side-by-side comparison images in Photoshop is that when the composite image was converted to sRGB from my usual ProPhoto RGB, the shadows darkened somewhat form my original adjustments.  I had never noticed that in prior work but that anomaly slightly changed the shadows in the comparison images.  Just FYI)

This first image below is 1 stop overexposed.  Let's start there.

1 stop overexposure, no adjustments (click to enlarge)
1 stop overexposed and corrected in Lightroom using only the Exposure slider set to -0.96,  That equates to turning
the exposure down about 1 stop to compensate, which seems normal. No other adjustments (click to enlarge)
Direct comparison of the nomimal exposure versus the corrected 1 stop overexposure.  I don't detect any noticeable
color shifts, no loss in shadow or highlight detail, no loss in texture or fine details in image. (click to enlarge)

2 stops overexposed, no adjustments made (click to enlarge)

2 stops overexposed and corrected in Lightroom by moving the Exposure slider to -1.83, highlights to -12
and clarity to +25 to bring out a bit more detail in the concrete.  A color shift of the sky to cyan but still
plenty of highlight detail.  Certainly a usable image if mistakes are made during exposure. (click to enlarge)


Here is a direct comparison of the above images (click to enlarge)
3 full stops overexposed, no adjustments made (click to enlarge)
You think that all is lost?  Look below.
This is a full 3 stops of overexposure and corrected in Lightroom only using the Exposure slider set to -2.98, which
is where you would expect it.  STILL no loss in highlight detail. Shadow detail and overall image detail is good.
There is now a noticeable color shift in the sky from blue to cyan, but that can be corrected easily in editing.  Also,
there is just starting to be a tiny bit of red color shift in the shadows but nothing significant.  This image is certainly
usable.  Think film.  Think color slide film.  More than 1/2 stop overexposure in a scene like this and the highlights
would have been lost forever.  Amazing!
Comparison of nominal exposure on the left and one that is a full 3 stops overexposed on the right (click to enlarge)
There is some color shift in the sky toward cyan.  However, there are still plenty of highlight detail.  Certainly an image
I could use for a variety of things.



This image is a full 3 1/3 stops overexposed.  No other adjustments (click to enlarge)
Not correctable?  Look below!

3 1/3 stop over exposed and corrected in Lightroom.  Exposure slider set for -3.08, highlights -27, shadows -37, and
clarity +38.  No other adjustments.
This is the limit for this sensor.  1/3 stop more and the highlights are gray and muddy and not correctable (click to enlarge)
There is starting to be some loss in highlight details in the concrete.  There is a large color shift to cyan in the sky.  Fine details are being lost and smearing is apparent in those small details. The shadows still look good and no more color shift in shadows.  This still is an amazing performance for any digital sensor, let alone a M4/3 sensor with 20mp.  This image,
in my opinion, could be usable if used on the web and kept relatively small. 
Now, let's go the other way.  Let's see what happens when we underexpose. 

Underexposure is much more desirable than overexposure and we know we have more latitude in correcting underexposures.  Instead of comparing -1, -2, etc. stops underexposed, which can easily be recovered and matched almost perfectly, let's start with a 4 stop underexposure and go from there.


Here again is the image at the top of this post with a nominal exposure in my judgment.  (click to enlarge)

4 stops underexposed. No adjustments to file. (click to enlarge)
4 stops underexposed and corrected in Lightroom using only the Exposure slider set to +4.04, which
is where you would expect a 4 stop underexposure to be.  I did increase the noise reduction to 25.  That takes
care of most of the noise and what is left, to me, is almost invisible and not objectionable except in the largest
prints or display.  No noticeable color shifts.  Good shadow detail.  It is hard to fully judge detail as I mentioned
above, the focus is just a bit off for this series of images.  However, it doesn't appear to be any change in detail. (click to enlarge)
5 full stops underexposed.  Is there an image in there?  If so, what kind? (click to enlarge)
Here is the 5 stop underexposed image corrected in Lightroom.  The Exposure slider was moved to +5 only. No
other corrections made.  No noise reduction added.  Fine details in the tree limbs is still visible in the 100% crop.  There
is still plenty of shadow detail.  There is a slight shift in the midtones toward green, but that can be corrected if desired.
The sky still holds pretty well.  The detail in the shadows is still visible. (click to enlarge)
This is the same 5 stop underexposed image as directly above.  In this image I attempted to correct the slight shift to green
in the midtones by adding 15 points of magenta. I then added 29 points of clarity, +27 to highlights and +31 to
shadows.  I then added some noise reduction, not fully however, to +21 as that seemed to be a compromise between
wiping out fine detail and reducing noise to a moderate amount.  It could go farther. (click to enlarge)
I think this image is usable if one were to post it on the web at a size no more than 1000 pixels.  Larger than that and
you can see the breakdown by noise of the tiny details in the branches.  The rest of the image survives this
extreme underexposure much better than those small branches.
Here is a direct comparison between the nominal exposure on left and one that is 5 stops underexposed on the right.
(click to enlarge)
Although it is not nearly as good as the nominal exposure, I find it amazing that the E-M1 Mark II's sensor can capture
and retain so much detail.  Again, think slide film.  A stop or two of underexposure and your shadow detail was
gone forever.
The 5 stops under is really pushing the limit of the sensor.  I can go 1/3 more stop to 5 1/3 stops underexposed and get the same result as above.  

Let's push the envelop a bit more and take a look at 6 stops underexposed!  You can pull out an image, but it is an image that could only be used under dire circumstances when you really needed something and didn't have anything else.  But it is there.  Again, amazing!


6 full stops underexposed and corrected as best I could in Lightroom CC.  There is still large detail in the shadows
but the small detail is totally obscured by digital noise.  Again, I added a bit of magenta to counteract the shift
to green in the midtones.  Also, I added a bit of clarity, highlights, shadows and whites to make it look as close
to the nominal exposure as I could. (click to enlarge)
There you have it.  A non-scientific practical application test of what you can expect from the Olympus OM-D E-M1 Mark II sensor.  I think the limits of usability are amazing, even for a digital camera.  The only camera that I have owned that has done a bit better, was my Nikon D810.   

In my view, I think the practical limits for the E-M1 Mark II are 3 1/3 stops overexposure and 5 1/3 stops underexposure.  However, to give yourself just a bit of better quality and room for error, I would keep the exposures between 3 stops over and 4 2/3 stops under. That is 8 2/3 stops of exposure latitude!

Obviously, it is optimum to precisely expose your image as that will give you the most flexibility in editing and the best quality.

So what does all this mean?  It means that the Olympus E-M1 Mark II is an amazing tool and, even if you are a pretty incompetent photographer and really over or underexpose an image because you have no clue as to how to basically operate a camera and refuse to set it on Program.  You can pretty much pull an usable image out of your hat to use even if you are that bad!  Additionally, you probably don't need to bracket exposures much for HDR images as the latitude built into this sensor along with the TruPic VII processor is kind of amazing!

If you see any flaws in my testing, please advise.  I try to logically think my way through these things but I can and do make mistakes.  You should do your own testing under circumstances you may encounter for your particular types of photography.  

Hope you have found this educational and interesting.
Thanks for looking. Enjoy! 

Dennis A. Mook 

All content on this blog is © 2013-2017 Dennis A. Mook. All Rights Reserved. Feel free to point to this blog from your website with full attribution. Permission may be granted for commercial use. Please contact Mr. Mook to discuss permission to reproduce the blog posts and/or images.

2 comments:

  1. Dennis,

    this is just a housekeeping note- don't think my earlier comment went thru. 3rd paragraph from end, refer to X-Trans Pro III sensor- that's Fuji :) Must still have that test on your mind! Thanks for doing this

    Rick

    ReplyDelete
    Replies
    1. Thank you Rick. These past two posts were a lot of work. I did some copy and paste with the narrative as there was no reason to write all of it all over again and I just missed that edit. Again, thanks!

      Delete