Event-Fotografie A. Tofahrn

Handling extreme light with DRI (Dynamic Range Increase)

In most situations you can't choose your light, and have to use what's available. Situations with extreme differences in light-intensity are a little problematic even for modern digital cameras. Most digital cameras have a lower dynamic range than analog film, and both won't even get close to the range of the human eye. Especially night shots are sometimes difficult to handle:
Workers and lamps highly overexposed. Environment underexposed, sky black.
With several different exposures of the same scenario, you can handle that problem with a procedure called DRI ("Dynamic Range Increase"). There are a several DRI-guides available in the net, one of them will be found below.
With a little experience results can be amazing:

Upon some request I decided to add one more DRI-guide to the public. Even though there are some tools available, which do that automatically, I still prefer doing DRI the hard way by hand, since you alwas keep control about parameters etc. With a little experience this is not that difficult either.
The DRI-process will help you to create attractive images even under extreme light conditions, when a normal exposed image will contain over- and/or underexposed parts. You're getting exactly these problems during night-shots, when the source of light steps into your frame.
I've choosed an image where even in that small resolution the effects of DRI are clearly visible:
Our human eyes handle such conditions pretty good. Digital cameras do not have such a huge dynamic range and ran into problems, when confronted with such. You normally have to decide, if you prefer correct exposure on the lamps or the water, for example:
Facades visible, water looses structure, lightcircle overexposed. Lightcircle and water acceptable, anything else too dark.
When you compare both images with the result above, you'll notice that the facades are taken from the brighter image, water and light came from the darker one. The tricky part is to combine these elements and make transistion nearly invisible. Depending on the source-images and your workflow this will succeed more or less.
To make DRI by hand you first need some congruent images from the same scene, which only differ in exposure-time (not in aperture!). How many pictures you need, highly depends on the light conditions and how your camera's sensor handle overexposures. My D70 has some blooming and bright pixels tend to flood their neighbors, the D200 handles them much more gracious.
For this example we're using only four images (D70, 12-24@12, f/9, 6s-0.7s, Whitebalance set to neon tube):
Then you'll need an imaging application which supports layers and transparent masks. Create a new project file and insert all your source-images as separate planes. The brightest layer should be on the bottom or background. In Photoshop this will look something like the snapshot left.
Begin your work with the lowest, the brightest layer. Disable all the other layers and select the brightest one. Now select all overexposed parts (highlights), in Photoshop there is a handy function already included (Select -> Color Range), which do that for you with a few mouse-clicks:
After accepting your selection, the selected areas will be marked as you see in the image above. Notice that in our example the fountain is selected but not the larger areas of the water. Since I also want to replace the surface of the water with a darker variant, I just use the mask-mode to add that surface to the mask manually.
The final selection in general has too sharp edges, especially when your sensor does some blooming around overexposed pixels. So we have to smooth the selection, what photoshop already has another handy function for. The tricky part is to specify the softness-factor in pixels (Sample radius). There is no table of correct values, as they depend on several individual factors like your images, what you want to achieve etc. You'll have to experiment a little, until you get the results you like, this is the art-factor within this process. As a rule of thumb you should know, that the width should be halved for each processed layer. In our example I've choosed 50, 25 and 12 pixels.
The final selection with the soft edge is now added to the next layer above as a mask, which effectively cut away anything outside your current selection. This is the part where darker areas are kept from the lower layers and their brighter parts being replaced by the layer above. You'll have to repeat this for each layer until you placed a mask for the highest layer. In Photoshop this will look like the snapshot on the left.
You can easily see, that the masks are getting sharper to the top and that I obviously added the water to both of the lower masks, so the water's surface is ultimately taken from "Layer 2".
In the final step all layers are reduced to the background and you can optimize the general impression with adjustment of contrast, brightness and saturation. Remember, DRI is more an art than realism and so you're free to form the result based upon your personal taste.
© 2004-2006, A. Tofahrn