A Digital Fill-Flash Technique for Improving Images

Bit by Bit: A Digital Fill-Flash Technique for Improving Images
My Backlit Pages
Using the example below, here‘s how to apply digital fill-flash to your images.
Theoriginal image of a marine naturalist in La Paz, Mexico, was taken inopen shade with strong backlighting (see figure 2). It does not showthe subject in the foreground adequately. I would like to modify it sothat the skin tones are lighter, while leaving the background untouched.
 

Figure 2
Making the mask.I examine the Red, Green and Blue channels, looking for the channelthat shows the best contrast between the foreground image and thebackground (see figure 3). In this image, the Blue channel shows agreater difference between the skin tones and the background, so thatis the channel I chose to duplicate to build my mask.
 

Figure 3
To duplicate a channel, I drag its title to the New Channel icon in the Channels window in Photoshop (see figure 4).
 

Figure 4
Acopy of the Blue channel is generated (see figure 5). I immediatelyrename the new channel "Mask" so there is no doubt what its purpose is(see figure 6).
 

Figure 5
 

Figure 6
Modifying the levels.Oncethe new channel is made, I modify it with Levels. This allows me toselect the subject, making it as dark as possible, while lightening thebackground and eliminating most of the tonality there. In this example,the sliders in the Levels control panel are tightly clustered in agroup that darkens the subject, and lightens the background effectively(see figure 7).
 

Figure 7
Thisis the result of the levels adjustments (see figure 8). I then Invertthe image (Command/Control-I) to create the negative of the mask image(see figure 9). Photoshop works with masks using the metaphor of filmmasks: white areas allow an effect to take place, while black areas aretreated as opaque, and prevent the effect from taking place.
 

Figure 8
 

Figure 9
Refining the mask.Once the image is inverted, I usually paint areas that I don‘t want tochange with the brush tool and black color. In this case I paint-out alittle bit of the background that shows up in the mask image. I followthat step in some cases with a very small amount of Gaussian Blur,usually between 0.25 and 1.5 pixels (see figure 10). This will softenthe edges of the mask, preventing its effect from being too obvious inthe final image (you don‘t have to be terribly careful here as thisprocess is very forgiving).
Figure 10
The final mask channel exhibits all of the characteristics needed for a selective tonal adjustment of the original image.
Adjusting the image with the mask channel.Once my Mask is complete, I return to the RGB view of the image byclicking on the RGB title in the Channels window. Everything is as itwas originally. I then go to the Select menu, and pull down to LoadSelection. A window pops-up, usually with Mask already selected. Ichoose my Mask channel and click OK, and my image assumes the selectionarea (with marching ants) based on my special channel. Since themarching ants are unpleasant to watch, I hide them withCommand/Control-H.
Tocontrol the quality of the shadows in my selection, I usually useCurves, but Levels will also work. A change of 15-20 percent in themidtones has a much subtler effect than the same change in an unmaskedimage. I always work in graphic arts terms in Curves (black in theupper-right, with percentage units) as opposed to the RGB terms.However, the same effect can be achieved in either mode.
Ilighten the image with Curves, until it looks much better, then clickOK. If you go too far, the image will have areas of posterization.
Theresulting image has significantly improved tonality, allowing thesubject to be seen better (see figure 11). If an overall tonaladjustment is needed, I can do that by deselecting the mask(Command/Control-D) and using Curves to further modify the entire image.

Figure 11
Ihave become so confident in this process that I usually throw my maskchannels away after I have completed the change, but if you areconcerned about the success of your work, you can save the image aseither TIFF or Photoshop with the extra channel intact.
Read more byBrian Lawler.