In addition to over 150 full-color images I've shot over the last six years, my new book, Window Seat: The Art of Digital Photography and Creative Thinking (O'Reilly Media, February 2006), also includes my suggestions for thinking creatively, as well as detailed information about how to make great-looking images of clouds, mountains, farmland, and water, while photographing through airplane windows. Here is an explanation of a few of the techniques I used to transform the images I captured with my camera into the final images that appear in the book.
A note on exposures: When capturing images, I usually try to open the aperture as far as I can. This allows me to focus on the clouds or landscape while minimizing the appearance of dirt, scratches, or ice on the multilayered window pane.
Figure 1-1. The final image
Figure 1-2. The original image, before processing with Photoshop
In this case, the original image was quite flat (lacking in contrast); it had a color cast in the highlights, and the left side was a lot lighter than the right (Figure 1-1). To create the final image, I had to do a lot of dodging and burning. The best way to do this is to use adjustment layers--they're flexible, non-destructive, and offer optimal control when making selective and global adjustments.
To begin, I selected the top and bottom corners of the left side of the image with the Lasso tool. I didn't worry about adding a feather to the selection because I knew that after I added the adjustment layer I would be able to soften the edge. I added a Curves adjustment layer (which automatically creates a mask from the selection). I pulled the curve down to darken the left side of the image until it matched the right. Then I blurred the mask of the adjustment layer (Filter -> Blur -> Gaussian Blur), changing the hard edge of the selection to a soft edge, so it was impossible to for anyone to tell which areas were adjusted.
Figure 1-3. The adjustment layers used to make corrections to the image
To brighten the highlights of the clouds, I selected them with the Magic Wand tool, created a Curves adjustment layer, and brightened them by moving the curve up. I blurred the adjustment layer's mask to soften the edges, as above. To remove the color cast in the highlights of the clouds, I chose Select -> Color Range, and sampled the highlights. Then I added a Hue/Saturation adjustment layer and decreased the saturation. To further decrease the saturation in the center of the image, I added a second Hue/Saturation adjustment layer and desaturated the entire image.
Since I didn't want the top clouds and bottom ocean to lose any saturation, I used the Gradient tool (set to Reflective), with the foreground color set to black and the background set to white, and drew a gradient from the center of the image to the bottom. This neutralized the saturation adjustment from the black areas at the top and bottom of the mask, while allowing the adjustment to affect the image in the center (where the mask was white). I repeated this process with a Curves adjustment layer to lighten the center of the image.
When I was done making my adjustments (Figure 1-3), I saved the file as a layered PSD file so that I could return to it at any time to make further changes if I wanted to. (Note: I add an "L" to the end of the filenames of layered PSD files so I can easily find the layered versions of the files on my hard drive.)
Figure 2-1. The final image
Figure 2-2. Original camera RAW image
Figure 2-3. Adobe Camera Raw settings used to process the original image
Because of the many layers between me (inside the plane) and the world outside, the images captured through the window are consistently flat. Shooting digitally compounds the problem--with analog, at least you can use a film (like Velvia) that boosts contrast so you don't have to make big adjustments to the original scan. On the other hand, correcting film that has already been processed and scanned has its limitations: for one thing, you're manipulating the second generation (the processed, scanned digital image, which is only as good as your scanner). With digital, you're processing the first generation--it's as if you're processing the original film itself, not an image scanned from film.
This image (Figure 2-1), which was taken on an overcast day, needed a lot of work. I used Adobe Camera Raw in Photoshop CS2 to process it as follows: I increased the exposure dramatically (+3.45) and moved the shadows as far as I could (+100), which started to really improve the image. I then took the brightness all the way down (to 0) and added contrast (+42). Because I made such radical changes to these settings, I also had to decrease the saturation (-88). I then used the Temperature and Tint sliders to make the images a more pleasing shade of green, which was also more in keeping with my memory of the scene. I turned off sharpening (0) and increased the Color Noise Reduction dramatically (100) to remove the artifacts that had been amplified by making such a big change to the tonal and color settings.
I brought the image into Photoshop in 16-bit at its native capture size (2000 by 3008 pixels) to avoid interpolation. To remove the dark spots (caused by dust on the sensor), I used the Healing brush as necessary. Finally, I added a Curves adjustment layer to increase the contrast and saved the file as a layered PSD file.
Figure 3-1. The final image
Figure 3-2. The original image. The white outline shows how I cropped it.
I knew when I shot this image that I only wanted to use part of it, and in a vertical format. The first thing I did was crop the file vertical, keeping the original aspect ratio of approximately 1:1.5 (Figure 3-2).
The intense blue cast and hazy appearance was due to fires in the area, and I wanted something less blue and less intense. I tried to adjust the color using Curves and Hue/Saturation adjustment layers, but was unable to achieve the look I was after. So I tried a different approach: desaturating the image and then adding color back in. I desaturated the image using the Channel Mixer. Because the Blue channel contained most of the image noise, I mostly used a combination of the Red and Green channels to successfully remove the color (Figure 3-3).
Figure 3-3. The desaturated image
After cleaning up the dust specks with the Healing brush, I added color with a Gradient Fill adjustment layer (set to linear) from purple to pink to yellow. The opacity of the layer was set to 50 percent, with the blend mode set to Overlay. This made it appear as if it the mountains were disappearing into the haze (Figure 3-4).
Figure 3-4. Adding false color with a Gradient adjustment layer
Finally, I dodged the top portion of the image by painting at a low opacity (5 percent) with white on an empty layer set to Soft Light mode (Figure 3-5).
Figure 3-5. The image with all of the adjustment layers active