There has been a lively and sometimes heated discussion on the luminous-landscape.com message boards about some of the techniques that can be used to vastly improve the level of shadow detail in an image while at the same time reducing shadow noise to almost nothing - essentially improving the signal to noise ratio throughout the image.
The author of the original post, Guillermo Luijk, introduced a sample program he had written which in its current state evaluates two input images, the first with the correct exposure for the highlights (U), the second four stops overexposed (O). Pixel by pixel, the program evaluates whether the pixel in O is blown out, and if not, adjusts the exposure to that of the pixel in U and replaces the pixel U with the exposure adjusted pixel from O. If the pixel in the O is blown out, it uses the pixel from U. The result is an image matching the exposure of U filled in with all the shadow detail and improved S/N ratio from O. The technique clearly produces fantastic results and I know he’s working to improve it and make it more flexible. Update: He has just posted an English translation of the original article here.
The one drawback to this technique is that it isn’t implemented (currently at least) as something I can easily use in my normal workflow which uses ACR and Photoshop CS3.
I’ve been experimenting with one of the other techniques presented in the thread that does fit in to my current workflow, and while it’s debatable which one produces the “best” results, the results of both are far superior to using neither. So here is a simple tutorial to get you started.
First, a little background about the image. The Montlake neighborhood in Seattle is home to the remnants of a 1960’s project to build a second North/South freeway that bypassed downtown Seattle (as opposed to I5 which runs through the city). The proposed interchange between SR-520 and the new freeway began construction but was halted when the project fell through. The partially built on/off ramps remain as part of the Washington Park Arboretum and offer a great photographic opportunities.
To start, here is the first exposure of the scene, 1/15th of a second exposure, known as Image1:
This image is basically properly exposed for the highlights but I knew I could overexpose slightly and use the Recovery slider in ACR to recover the highlights.
Here is the second image, 1/4 of a second exposure, Exposure slider -0.25, Recovery slider +100, known as Image2:
At this point I have pretty much the best image I could get using a single exposure and the available tools while maintaining detail in the highlights. The histogram looks similar to the first image but there is less clipping in the shadows.
In order to show the difference in the shadow detail between the two I’ll apply this curve (yes, it’s extreme but this is an example!):
Here is a closeup of the shadow area under the structure with the curve applied.
You can see that in Image1 the detail is minimal, almost completely obscured by noise, and it’s just plain ugly. Image2 benefits from the slight overexposure and handles such an extreme adjustment much more gracefully, however there is still quite a bit of noise. Image2 is an example of the Expose to the Right technique (ETTR if you’re searching for it online). But that’s not all!
I also made a third exposure at 1 second - approximately 3 stops more than Image2, known as Image3:
Clearly, if the concern was maintaining the highlights then this image is severely overexposed but the goal was to capture the shadows at a level which can be blended in to either Image1 or Image2 to improve the shadow detail and to make our final adjustments more flexible.
My goal was to maintain the look of Image2 while being able to make some extreme adjustments in the shadow areas to bring out the structural elements underneath the bridge. In order to do that, I adjust Image3 in ACR with a -2.8 exposure (approximately the amount it was overexposed to begin with). The amount is arbitrary really but I wanted to match the look of Image2.
Here is the resulting Image3 with -2.8 exposure:
The shadow areas are pretty close to Image2 and the highlight areas are still blown out.
The goal now is to merge the high signal to noise ratio shadows/mids from Image3 in to Image2. To do so, follow these steps:
1. Layer Image3 on top of Image2 in Photoshop.
2. Right-click Image3 and select ‘Blending Options…’
3. Using the ‘This Layer’ Blend-if slider, move the white slider to 235, then alt-click the slider to split it and move the left half of the slider to 35. Again, these numbers are approximate and you can adjust accordingly if necessary.
Looking at the result in Photoshop is nothing spectacular - it looks just like Image2 in the bottom layer. The real benefit is when we re-apply that curves adjustment. Here is the result of blending Image2 and Image3 with the curves adjustment (Image2 alone with the curve is shown again below for comparison purposes):
The resulting image, even after the curves adjustment shows, for all intents and purposes, zero noise and as much detail as you could ask for (from my camera at least!) - even in the deepest shadows. You can see the difference more clearly by turning the visibility of the layer with Image3 on and off. Alternatively, if Image2 wasn’t available, the blending with Image1 is even more dramatic and would have allowed you to actually use Image1 instead of quite possibly discarding it.
The key point here is that after creating the “improved” image you can introduce huge tonal shifts to shadow areas of the image with little to no degradation in quality - effectively expanding the potential to reach the intended look or vision for the image.
Lastly, a disclaimer. I am not claiming to be the first person to use this technique nor the first person to mention it online. I also do not claim that there is no other way to accomplish the same thing - like HDR for example. My only goal here is to illustrate what a dramatic difference can be had in image quality by taking steps at the time of exposure, when it’s possible of course, to ensure you have the best “base” image to work with in the digital darkroom.