Friday, October 11, 2013

Recovering from Underexposure - A New Approach

In my last post, I discussed my strategy for exposure for digital cameras, which usually results in images that are underexposed when downloaded straight-out-of-the-camera (SOOC).  In this post, I would like to discuss my current strategy for recovering shadows.  Here I use Lightroom but the same techniques can be used in other post-processing software.


Here is the SOOC version of the image above.

I used the approach I discussed in my last post, relying on ADL to expose for the highlights.  As expected, the SOOC is underexposed.  However, the relevant highlights are preserved:

even the candle highlights are preserved

The highlights in the window on camera right are not preserved but that's ok.  They don't add to the image (it was an overcast day with gray skies) and exposing for them would have led to more noise without much offsetting benefit.

Easy approach

The  easiest approach is of course to use the shadow slider (increase it as needed) and the highlight recovery slider (decrease as needed).  In my opinion it is also the least satisfying way to recover underexposure.  The result looks HDR-ish (but if you like that effect, then this is perfect for you -- boost the clarity slider next to make it look more like HDR).  Moreover, in some sensors like that of the D7100, using the shadow slider is more likely to reveal noise or worse, banding.
Using the shadow slider: HDR much?

Global approach
The next approach is one that I've posted about before.  In brief, instead of relying on the shadow slider, I use the exposure slider.  I do use the shadow slider as well, but only up to about +20 or sometimes +30.  More than that and it looks too artificial for me.  I use as little of the shadow slider as possible.

A +20 or +30 shadow adjustment for a scene that's exposed for highlights will nowhere be enough.  Therefore the heavy lifting ( sorry :) ) is done with the exposure slider which I use as much as necessary.  Using the exposure slider will be more likely to blow the highlights that are actually captured.  To offset that, I use aggressive highlight recovery.  As much as -200 (by using an adjustment brush on top of the highlight adjustment).  Unlike using the shadow recovery slider I don't find any issues with the highlight recovery.

For very contrasty scenes I get a little help by changing the camera calibration to a low contrast picture style and if necessary, by decreasing the contrast as well.

When the midtones look normal, I then work on restoring the oomph because the image will otherwise look very flat / low contrast.  To do that, I increase the white slider (making whites appear more white), and I decrease the shadow slider (deepening the blacks).  Does that effectively negate what I did with the highlight recovery and shadow recovery?  Not at all.  The net effect of those changes is to move the shadows and highlights a little closer to the middle, where it is easier to see them (partly because our eyes can differentiate more midtone shades).

Here is the result:

Here again is the first method for comparison:

This second method produces a more natural and more three-dimensional appearance.  Now let's talk about the third approach.


I call my current approach the Hobby approach for two reasons: first, it is styled after Strobist David Hobby's logic of using lights; second, because it is customized for each image, it is probably not usable for event pros who need to edit thousands of shots.  So, it's just for us hobbyists. :)  Well, ok, editorial and portrait shooters too.

Until about a year ago, I was a huge strobist fan.  My lighting collection attests to that (someday I'll review more of my lighting gear).  I'm still a fan, but after having gotten the D600 with its exceptional shadow recovery abilities, I found that I could get results with postprocessing that closely mimic real artificial lighting (lol).  Here's how it works (strobists among you will recognize the logic):

Shadow detail legibility
The first thing I do with the underexposed image is to raise the exposure for shadow legibility.  In strobist-speak that means increasing the exposure to the point where the details of relevant shadows can be seen.  What that exactly means varies depending on your shot and your taste.  If for example you want an intense-looking dramatic scene then the shadow details would be barely visible.  If you want a look that mimics Mr. Hobby's shooting style, you want shadow details that are pretty easy to see (enough that if you print the shot on newsprint with a lousy printer, you will still have enough detail).  I myself prefer something in between.

Here again is the SOOC shot:

And here is the shot with the exposure raised to my desired shadow legibility (in this case +0.36):

To tame the highlights, I use the highlight recovery slider, that's all.  I don't mind being aggressive with it, although I watch to make sure the image doesn't become too flat.  For stubborn highlights, I use an adjustment brush or radial gradient to bring them down further.

To raise the midtones, I use Lightroom 5's radial gradient.  To me it looks more convincing than using the adjustment brush.  If you're new to LR5, pls. note that the radial gradient can be changed to an oval/ellipse and can be tilted.

I use an inverse mask which means the effect will be within the circle that you apply instead of outside it.  I also increase the feathering from 50 to 75.  This simulates real world lights more closely (unless you want it to look like you used a snoot, in which case you decrease the feathering).  For each gradient I try not to increase more than +0.7 or else the result looks too obvious.

I apply this radial gradient everywhere I want to raise the midtones, starting from the darkest midtones.  For some areas, I apply overlapping radial gradients that are progressively smaller.  For example if I have a portrait shot from the waist up then I'll apply a radial gradient for the head and body, then to highlight the face, I'll add another radial gradient just to the head area.  Note that I don't just use one huge radial gradient to cover the arms as well -- I apply a separate radial gradient to each arm.

From the previous shadow legibility step:

Here are the areas where I applied a radial gradient (see the dots):

Result with midtone recovery:

As with the second approach, I add punch to the image using the white slider, black slider, and sometimes the clarity slider.

That's pretty much all there is to this approach.  Note that this method cannot recreate rim lights or kickers or hard lights.  For those I use ambient light sources or flash.

Here are some more samples: