Un-debunking ETTR

On a recent thread on Digital Darkroom, a post was made in an attemp to "Debunk ETTR" which stands for Expose To The Right when shooting raw digital captures. The ETTR approach was written about by Micheal Reichmann of the Luminous Landscape web site and came from a discussion he had with Thomas Knoll the coauthor of Photoshop and founding engineer of Camera Raw. The theory goes that IF the contrast range of the scene you are shooting is less than the dynamic range of your camera's sensor, you can increase the quality of the capture by adjusting your image to be further to the right on your camera's histogram display. You do this by either opening your lens, slowing down the shutter.

This is not intended to be used in every photographic situation, only those situations where the scene's contrast range is lower than you camera's sensor dynamic range. Clearly if you have highlights whose texture and detail is important, you wouldn't want to increase the exposure to the point where the highlights are clipping. If you can spare the increase of the F-stop or the lowering of your shutter speed, increasing the exposure can decrease the noise found in the shadows of digital capture. The reason for this isn't magic, it's simply that the more photons that fall on your sensor, the better the quality of the capture and the better the signal to noise ratio.

Below is a typical scene where ETTR can help.

These captures where shot on a Canon EOS 1Ds MIII set to auto bracket three exposures. The histograms from Camera Raw show that there is shadow clipping in the -1.33 stop exposure as well as the normal exposure. The +1.33 stop exposure shows a clipping indicator but the question is, is the extreme highlight data clipped or merely clumped way up at the top? As you can see, this is not a particularly low contrast scene, but will the +1.33 stop be able to be controld to not clip any data?

These are the same three images after adjustments were made in Camera Raw. Note that all three images have been adjusted. I felt the original normal exposure needed to be brightend up. I was able to get each of the three shots to be very, very similar in tone and color. You can see that the resulting histogram while not identicle are quite close. At this size on the screen, all three captures look decent. But, is there a difference when looking close?


Click on the image to open the full 1800 pixel wide comparison at a 300% Photoshop zoom.

What can you take aways from this series of images? First off, it's really bad for image quality to under expose an image and try to get the image lighter after the fact in post processing. While you can make the image look "better" under exposing substantially reduces the signle to noise ratio. Even the "normal" exposure has a degree of noise that while isn't ubjectionable, will potentially imact how much sharpening you can apply and how big the image could be reproduced in print. Clearly, looking at the image at 300% in Photoshop won't tell you what the image will look like after going through a down-sample, but there is no question that the +1.33 exposure produced a cleaner, less noisy and better signal to noise ratio.

But, what about the "risk" that by increasing your exposure you may blow out your highlights? While it's certainly true that highlights can be clipped and loose all detail if you over expose, that's not what is being advocated here. If the scene contrast range is lower than the dynamic range of the sensor, increasing the exposure just short of clipping won't put your image at risk. And, if there is any question, you can bracket–if the shot you are doing will give you enough time.

On the other hand, most digital cameras can capture a ton of highlight detail in a raw capture. While the camera may show a clipping warning, the fact is the camera is designed to be conservative. The following two images are an attempt to prove just how much usable detail you can tease out of an image Exposed To The Right (WAY to the right).

The image above was a shot of Niagara Falls in an attempt to shoot with a real slow shutter speed in no-sun daylight. But since the lens was already stopped down all the way and the ISO was at the lowest setting, doing a "bracket" of slow shutter speeds resulted in what I thought was a lost cause. An image whose data was clumped way to the right–and extreme case of ETTR.

This is the result of pulling the data way down on the tone scale. Ironically, Highlight Recovery was about 45 and Exposure was at -0.08. The main adjustment was to set Blacks to 100 and adjust the Shadows of the Parametric Curves also to -100.

What is this designed to prove?

That there's a heck of a lot of usable data in a raw capture.

And ETTR isn't "risky" if you take care and know what you are doing...

 

To my main web site.