Unlocking HDR; Examining Photoshop’s HDR Mode: Are There Better Ways To Deal With High Dynamic Range Images? Page 2

These HDR images require a special 32-bit file format with a greater brightness range than any monitor or printer could reproduce, and which must be converted into regular 8-bit or 16-bit files before they can be used. The conversion option you choose governs the appearance of the finished picture. "Exposure and Gamma" gives results closest to what you might expect from a negative with a very high dynamic range, but yields predictably low contrast. "Highlight Compression" (seen here) preserves more mid-tone contrast but leaves highlights looking flat, and "Equalize Histogram" produces better contrast in some areas but with even more pronounced "dead spots" in others. "Local Adaptation" is an attempt to blend detail from both exposures, but the results are crude and artificial-looking compared to the conventional layering techniques we examine next.

The Layer Blend Approach
Here, our two "exposures" have been superimposed as layers in Photoshop, with the "shadows" layer on top. The blown highlights in this top layer are selected with the Magic Wand tool, and this selection is inverted and converted into a layer mask. Now, the highlight detail preserved in the lower layer shows through, though the edges of the mask will be too sharp, and the "join" will look artificial. The solution is to apply a Gaussian Blur to the layer mask. A value of 250 pixels gives a much subtler blend. Our picture is still not quite right, though. It's possible to see that the bright background hillside seen through the arch grows subtly lighter toward the edges of the arch, and this is because the blurred layer mask now extends into this area.

This effect may not be obvious, but if it is the solution it involves just one extra step in the process. Once the initial selection is made with the Magic Wand, use the Select>Modify>Expand command to expand the selection by, say, 50 pixels before inverting it and creating the layer mask. Now, the "blend" in the mask is shifted to overlap with the darker areas in the arch, where it will be noticed less. The improvement in this version of the image is subtle, but useful. (It's possible to see the layer mask is slightly larger and denser as a result, too.)

The way this layered image is constructed provides a great deal of editing flexibility. For example, we've been able to enhance the contrast of the background hillside and the interior independently, with curves adjustments--remember, they're on separate layers and the layer mask prevents adjustments in one having an effect on the other. Also, the inside edge of the arch on the left looked a little "dead." This is caused by the blurring of the mask in that area. The solution was to paint out this part of the mask using a small, soft brush (70 pixels) with the color set to white. This "brushed away" the mask in that area so that the detail in the brighter "highlight" layer became visible again. These manual touchups to the layer mask are a quick way to iron out any remaining problem areas.