Creating High-Dynamic-Range Images

By now, you've probably been doing some raw work of your own and have experimented to see the difference between 8-bit and 16-bit images. When you tell Camera Raw to produce a 16-bit image instead of an 8-bit image, you should be able to "feel" a difference in the resulting files when you start to edit them. With 16 bits of data for each pixel, a 16-bit image can contain a much larger variety of colors than you can get in an 8-bit image.

The range from black to white may be exactly the same whether you're storing in 8 bits or 16 bits, but a 16-bit image will have a much finer gradation of tones between those two points. This finer gradation means that you can make finer edits without worrying about creating tone breaks.

In the previous layer masking example, though, our problem was not a lack of intermediate tones. Rather, it was that the camera couldn't capture the extreme range of light to dark that was in our scene. As discussed earlier, a camera can capture only about 8 to 10 f-stops (or EV) of light, while some outdoor scenes can have a range of 12 EV. So in the previous example, the only way to capture the full range of light that was present in the scene was to shoot two exposures and composite them. Our first exposure captured the bright areas of the image, and our second captured the dark areas. By compositing them, we effectively created an image with a much wider dynamic range than a camera can capture in a single shot.

NOTE

This dynamic-range limitation is not just a digital limitation. Most films have about the same range, though some black-and-white films have a slightly greater range, but still far less than your eye.

With Photoshop CS2, Adobe added support for 32-bit images. As its name suggests, a 32-bit image stores 32 bits of data per pixel. With 32 bits, you can store a mind-numbing variety of colors. However, 32-bit image mode is not intended for everyday use. Hardly any of Photoshop's regular tools or commands work in 32-bit mode, and even if they did, they would be tedious to use. A 32-bit image contains such a tremendous amount of data that even a very speedy computer would grind to a halt if you tried to perform the processor-intensive tasks that you normally execute in Photoshop.

Adobe added 32-bit support to facilitate the creation of high-dynamic-range (HDR) images. When you have 32 bits of storage space per pixel, you can represent a much larger dynamic range than when you have only 8 or 16 bits per pixel. HDR provides a way to create images that have a much higher dynamic range than what a camera can capture in a single exposure, and it allows you to do this without any compositing or masking.

HDR basics

To create an HDR image, you first need a scene with more dynamic range than your camera can handle. Creating HDR images is a bit of a chore, so HDR is not something you want to use for everyday shots. In addition to the hassle of carrying a tripod, you'll need to spend some time properly shooting your scene. You'll then spend a fair amount of time in Photoshop putting your final image together.

The most common HDR application is landscape photography, where it can be very difficult to properly expose a foreground and sky in the same shot. Interiors of buildings can also benefit from HDR when you want to capture both a well-exposed interior and the details outside a brightly lit window.

As with shooting a composite, you make an HDR image by taking a series of pictures of your scene, each with a slightly different exposure (Figure 6.52). We'll discuss the specifics of HDR shooting in the next chapter, but your goal is to shoot at least three images (and preferably five or six) with varying exposures.

Figure 6.52. For these seven images, I increased the exposure by 1 stop for each shot. These seven exposures will be used to create a high-dynamic-range 32-bit image.

Your intent is to produce a series of images that fully reproduce the dynamic range of the sceneso you'll have underexposed images that properly capture the bright areas of your scene, and overexposed images that properly capture the shadow areas of your scene. This is just what we had in the previous compositing example, but with HDR, you have more source images and no masks.

Camera Raw processes your source pictures and then hands them off to Photoshop, which analyzes them. Photoshop pulls different ranges of brightness information from each of the exposures and then combines the pictures to create one 32-bit image with the full range of bright to dark that you captured.

One of the most impressive things about Photoshop's HDR capability is that it does a fairly good job of mimicking some of the things that your eye does. In the process of combining and merging your HDR source images, Photoshop compresses the overall contrast range of those images without reducing the contrast of any well-defined edges in your scene.

Creating an HDR image

Obviously, shooting your source images is the first step in creating an HDR image, and you'll learn how to do this in the next chapter. Once you have your sources, you're ready to begin merging.

Merge to HDR

You can launch the Merge to HDR feature from either Bridge or Photoshop. For this example, we'll be hosting the entire procedure in Photoshop.

In Photoshop CS2, select File > Automate > Merge to HDR. In the Merge to HDR dialog box (Figure 6.53), you select your raw source files. If you want, you can convert your source images yourself using Camera Raw and then open them in Photoshop. Set the Use menu to Open Files, and it will work with the currently opened images.

Figure 6.53. The Merge to HDR dialog box lets you select the source images that you want to merge into a final HDR image.

When you click the OK button in the Merge to HDR dialog box, Photoshop will spend some time opening all of your source images and then will present you with a new window containing your merged composite (Figure 6.54).

Figure 6.54. After performing the merge, Photoshop gives you a chance to deactivate the source images and to view the entire 32-bit range of data that you have collected.

Don't be disappointed

Here's a warning: If you thought that an HDR image would have a breathtaking sense of color and realism thanks to its much greater dynamic rangeif you thought that you were going to see an image far beyond what you've ever seen before from your digital camerathen you're going to be a little disappointed.

While your 32-bit image may be overflowing with useful image data, your computer monitor is still, at best, a 10-bit display device. It simply can't come close to showing the huge range that may be present in your new 32-bit image. Sadly, a reflection print on white photo paper offers a dynamic range of only about 100:1, or 8 bits.

So what's the point of going to all of this trouble if you can't see the data on your monitor or printer anyway?

Think back to our compositing example from earlier. Our final image did not have all of the information that was contained in our two source images. We culled some tonal information from one image and other tonal information from the other image, and in some parts of the image, we blended tonal information from both sources.

HDR works kind of the same way. Though you can't see all of the tonal information that's present in your 32-bit image, the data is there. Where the data becomes useful is when you use some special conversion tools to turn your 32-bit image into a 16-bit image that contains tonal values culled from different parts of your entire range of source images.

Explore the data

The resulting image shown in the Merge to HDR window may look a little strange. First, notice that it has more shadow and highlight detail than any of your individual source images. However, it may appear a little washed out, and some highlights may even appear to be blown.

The left side of the window shows a scrolling thumbnail display of all of your source images, along with the exposure values that Photoshop calculated for each image. For example, in Figure 6.54, the first image has an exposure value of +3.32, indicating that it was overexposed by 3.32 EV.

The check boxes allow you to turn off each image. Deactivating an image removes its data from the merge. This can be an interesting way to see which data Photoshop is pulling from which image. Remember that the brightest image in your sequence is contributing the darkest information. In this case, the bright image is contributing a lot of the darker shadow tones near the base of the columns. Conversely, the darkest source images are contributing the bright areas of the sky.

Based on what you've been learning, this is good news. Since the shadows are coming from the brightest images, the ones that are overexposed, then they'll most likely be noise free.

On the right side of the window is a Set White Point Preview slider. This slider does not affect any of your image data. It's there only to help you explore the full range of data in your merge. By moving it, you can view brighter or darker parts of your 32-bit file.

Though the Set White Point Preview slider doesn't change any of your actual image data, its setting is stored with your 32-bit image and is used by Photoshop later in determining what the image should look like on the screen, so be sure to leave it on a setting that presents a pleasing image.

Create the merge

Click OK, and Photoshop will perform the final merge and deliver a 32-bit file in a normal Photoshop window (Figure 6.55). If you poke around in the Photoshop tool palette, you'll find that most of the tools aren't functional. However, you can use Photoshop's Rubber Stamp (Clone) tool, which is good for making minor touch-ups. You can also sharpen, blur, and apply a few other filters as well as resize and rotate the image. For the most part, though, you won't be doing any serious editing on the 32-bit file. In this case, my merge suffered from some troubles with camera sensor dust, so I removed these with the Rubber Stamp tool.

Figure 6.55. After the merge, you'll have what looks like a normal Photoshop file. If you look in the title bar, though, you'll see that it's a 32-bit image. Because most Photoshop tools don't work on 32-bit images, there's very little you can do in the way of editing at this point.

Save the image

Before you can start any serious editing, you need to convert your image to a 16-bit file. Just as your raw file can be processed in different ways, you can perform different 16-bit conversions of your 32-bit image, so it's a good idea to save the 32-bit file so that you can try different approaches to the conversion. Choose File > Save and save the image in Photoshop format.

Convert to 16-bit

Select Image > Mode > 16 Bits/Channel to invoke the HDR Conversion dialog box. This is where the real work of making a good-looking HDR image takes place. To perform the conversion, Photoshop has to figure out which tonal values to throw away to fit the huge mass of 32-bit data into a 16-bit data space. To allow you some control of the process, the HDR Conversion dialog box presents several conversion options, each of which provides different controls (Figure 6.56).

Figure 6.56. Photoshop provides several methods for converting your 32-bit image into a more usable 16-bit image.

The default setting is Exposure and Gamma, which provides two sliders. Exposure makes an overall brightness change, and Gamma shifts the brightness of the midpoint. For an image with a lot of range, this is kind of a brute-force tool that may not allow you to achieve the contrast that you want, especially in the darker parts of your image.

Highlight Compression simply compresses the brightest values in the 32-bit HDR composite, so that they fit within the range provided by a 16-bit image. While this conversion brings lots of shadow detail out of your image, it can be really brutal on highlight areas.

Equalize Histogram tries to compress both shadows and highlights to fit the whole HDR range into 16-bits. This conversion is easier on your highlights, but it may cost you a lot of shadow detail.

For images that don't possess a huge range, Highlight Compression and Equalize Histogram can work well. For most images, though, you'll want to use Local Adaptation. Local adaptation provides a curve interface that lets you create a tone curve that alters the brightness of specific parts of your image.

When you first select Local Adaptation, things may look pretty grim. You'll need to click the arrow to open the Toning Curve and Histogram display (Figure 6.57).

Figure 6.57. Local Adaptation is your best option for making 16-bit conversions from 32-bit images. Though the image may look bad at first, the editable curve provides a tremendous amount of control.

Just like the curve in Camera Raw or Photoshop, the Local Adaptation curve lets you roll over your image to identify where specific tones are represented on the curve. Command-clicking automatically places a control point on the appropriate spot of the curve.

Your goal is to place control points on your curve and position them to set the brightness levels of different tonal values in the image, just as you would when editing a curve in Photoshop. If you're not sure where to start, try following the histogram that's superimposed behind the curve. Where there's no data in the histogram, there's no need for a curve adjustment. Anchor the curve where the data begins and then start your adjustments by reshaping the curve into an S-shape (Figure 6.58).

Figure 6.58. A good place to begin your editing is to set up a simple contrast-inducing S-curve. Note that most of your curve adjustments will be extremely subtle.

This will give you a reasonable amount of contrast and at least make the image appear a little normal. From there, you can sample points and make adjustments to bring out more or fewer shadows in some areas and to improve the contrast and brightness of the midtones and highlights in other areas (Figure 6.59).

Figure 6.59. After a few more control point additions, this curve works well for our image.

One important difference between the Local Adaptation curve and the other curve controls you've used is that the Local Adaptation curve allows you to create corner points. Select a point on your curve and select the Corner Point check box beneath the curve display, and your point will become a corner, allowing you to isolate that point of the curve from the effects of adjacent control points.

Above the curve control are two sliders: Radius and Threshold. These control the way that the curve treats the varying brightness values of neighboring pixels. To understand how these sliders affect your image, you need to play with them. In general, the default settings are usually best.

Even slight adjustments to the curve can have a big effect on your image, and you'll have many choices about how to adjust different tones. Figure 6.60 shows a different Local Adaptation conversion of the same image that was used to create Figure 6.59.

Figure 6.60. Here's the same image data converted using a different Local Adaptation curve. Because a 32-bit image contains so much information, you can create very different interpretations of your scene simply by changing the conversion curve.

Using the Load and Save buttons in the Local Adaptation dialog box, you can save different curves, allowing you to experiment with different conversions of your data. As you can see, by simply reshaping your curve, you can create a different interpretation of your image.

NOTE

If you want to go directly to 8-bit from your merged 32-bit image, you can simply choose Image > Mode > 8 bits/Channel. You'll be presented with the same conversion dialog box, but your resulting file will be an 8-bit image.

Click OK, and Photoshop will perform your specified 16-bit conversion, and you'll have a normal, editable 16-bit image. To give you an idea of how much more dynamic range you can capture with an HDR image, Figure 6.61 shows the best single image that I was able to capture compared to my final HDR composite. The HDR image has much more detail in the shadow areas and sky, and much better tone in the midtones, offering slightly better color.

Figure 6.61. Here you can see the best single exposure that I could get of this scene (left) alongside my final HDR image (right). In addition to the improved shadows and sky in the HDR image, notice the more even tone on the midtone areas of the columns.

Refining an HDR image

Although you can do a lot of tonal correction during your conversion from 32-bit data, your image will most likely need a little editing once you've arrived at a usable 16-bit image.

Correcting motion

The trees in the upper-right corner of the final image have some annoying magenta fringe around them (Figure 6.62). This is not chromatic aberration, but an unfortunate result of the fact that the trees were blowing in the wind while I was shooting my separate exposures. Because they're not properly registered from shot to shot, they are slightly blurry and fringed with color in the final merge.

Figure 6.62. This annoying magenta fringe is the result of the trees blowing while I was shooting my separate shots. Because the trees are not properly registered, they don't merge correctly.

There's no simple way to eliminate this fringe, so rather than try to correct it, we'll simply replace the trees. We'll begin by using the Rubber Stamp tool to remove the original trees from the scene (Figure 6.63).

Figure 6.63. Rather than trying to remove the magenta fringe, we're going to simply replace the tree. Using Photoshop's Rubber Stamp tool, we first erase the trees in the merged image.

Because they're heavily backlit, the trees look fine as a silhouette. Find a source image that has a well-exposed sky and use Photoshop's Rectangular Marquee tool to select the corner of the image where the trees are. Then copy this selection and paste it into the merged document (Figure 6.64).

Figure 6.64. The tree image was copied from one of the original source files and then pasted into the final merge.

Since we picked an image with a well-exposed sky, we have a strong blue background behind the silhouetted tree. Eliminating it is very simple. With the Magic Wand tool, click the sky to make a selection and then choose Select > Similar to select all of the other sky tones in the pasted layer. Then choose Select > Feather and add a 1-pixel feather to the selection, to create a smoother edge around the selected area.

Pressing Delete deletes the selected sky, leaving the tree silhouetted against the HDR sky. Because the selection we pasted in was rectangular, it included more than just the trees in the corner. A section of the columns was also pasted, but we can remove it easily with the Eraser tool (Figure 6.65).

Figure 6.65. After deleting the sky and the extra parts of the building that were pasted in, we're left with an original tree silhouetted in front of the high-dynamic-range sky.

In general, it's best to pay attention to this sort of thing while shooting. If you're in a windy location, don't try to focus too closely on moving objects.

(While removing the trees, I also took the opportunity to remove the flood light from the top of the structure and a plastic cup that some loser had left at the bottom of one of the pillars instead of putting in a trash can.)

Correcting tone

Unfortunately, even after all that fancy high-bit work and conversion, your image may still need some tonal corrections using Photoshop's regular toning controls. To create the final image, use adjustment layers to improve the contrast of the scene and to add a little localized saturation.

Because we use layer masks for these edits, you may be thinking, "Oh great, we're back to mask painting? So what was the point of all this extra trouble?" But our masks were very simple, and also, because every part of the image has much more detail in it, we have a file that allows much more manipulation than the sandwiched composite that we looked at before. And because the shadow details were pulled from the brighter images in the sequence, we have a final image that is noise free from its brightest to darkest tones. Figure 6.66 shows the final result.

Figure 6.66. The final image after it's been retouched and had a slight levels and saturation adjustment applied and a little bit of sharpening.

Too real?

One of the most curious things about high-dynamic-range images is that they don't always look "right," and one of the reasons is because they don't look like the photos we're used to seeing. We're accustomed to seeing photos with a very limited dynamic range, which often means an image with very dark shadows or very bright highlights. An HDR image, with its tremendous range, often lacks these extremes and so can look a little strange to us.

Photography is a representational medium, and normal photos are abstracted quite a way from reality. HDR, with its expanded range, creates images that are more like what your eye actually sees, but fairly far from what you're accustomed to seeing in a photo. In addition, abstracted images are often more appealing, because they force your imagination to fill in details on its own. When you provide more detail in the print, there's less for the viewer to imagine, and the viewer has less personal involvement in the viewing process.

When working with HDR images, therefore, you have to make a number of decisions about how much extra range you want to reveal. As you can see from the Highlight Compression sample in Figure 6.56, simply adjusting your image so that every tone is visible does not create a very good picture. You may even find yourself obscuring some shadow detail to make a more pleasantly contrasty image. What's nice about HDR is that it provides a tremendous amount of latitude for making these types of decisions.