Processing HDR images involves three stages. The first is merging a sequence of source images into a single HDR image, that in turn involves automatic alignment, rotation, or cropping of the source images. The program searches for correctly exposed image areas and ignores over- or underexposed portions of the frame. Every HDR program has its own methods for performing these steps.

The resulting image can usually no longer be stored as a 16-bit image, but rather must be stored (at least temporarily) in one of the currently popular HDR formats, such as OpenEXR, Radiance, or extended 32-bit TIFF.

Such high-color-depth images can not be effectively displayed on conventional monitors, be printed on everyday printers, or be edited using conventional image processing software.

Figure 9-38. Producing HDRI images is a three-stage process.

The second phase of HDR image production therefore involves reducing the image’s tonal range to a viewable (or printable) format without losing any of the improved detail. This process is called Tone Mapping, and there is no single, ideal way to tone map an HDR image. Most HDR programs offer various tone mapping methods, and the better the method, the more adjustment parameters it will have. Tone mapping is by far the most complex of the three phases, and offers you space for exercising your creativity and producing highly varied versions of the same basic image.

The third step in the process is the familiar image optimization step. It is often necessary to selectively dodge or burn specific parts of a finished HDR image, or to adjust the colors to more realistic levels.

There are a number of high-quality HDR programs available, including freeware such as FDRTools Basic [72] or the open source Qtpfsgui.^[134] Most commercial programs are available for between US50 and US$100. Each solution has its own strengths and weaknesses, and most are in a state of constant development.

The most well-known program on today’s market is HDRSoft’s Photomatix [77]. The program’s particular strengths are its great basic functionality and its high-quality tone mapping. There is an example of what Photomatix can do in HDR Imaging Using Photomatix Pro. We also like Andreas Schömann’s FDRTools Advanced [72]. This is a great value shareware program and includes a tone mapping plug-in for use with Photoshop. It is the only program we know that allows you to make extensive personalized settings during the HDR production process. We also recommend HDR Photostudio [105] (and its follow-up version HDR Expose) and PhotoAcute [81], which can be used to quickly create simple HDR images using just a few adjustable settings.

Photoshop, too, includes HDR functionality since the CS2 version. Its functionality is constantly being improved and is especially effective in CS5. We nevertheless only use Photoshop for processing scenes with relatively low dynamic range.

Although Photoshop’s tone mapping functionality is not as good as that offered by Photomatix or FDRTools, the basic HDRI production module is very effective, especially when it comes to automatic alignment of source images or processing moving objects within the frame. We sometimes use Photoshop to produce a basic HDR image before saving it for tone mapping with a plug-in or a different, standalone program.

When saving our HDR image using File ▸ Save As, we have a number of different formats at our disposal (Figure 9-42), ranging from standard Photoshop PSD to specialized 32-bit TIFF. Each of these formats has specific advantages and disadvantages. A detailed explanation would be beyond the scope of this book, but Christian Bloch’s HDRI book [11] provides comprehensive information on the subject.

Figure 9-42. Photoshop (here CS3) offers a number of different HDR output formats.

If you want to use a standalone program to tone map your image, we recommend that you first save it to OpenEXR format (with the “.exr” filename extension) or to the Radiance format (with the “.hdr” extension).

HDR images can be optimized using Photoshop, but not all of the typical Photoshop functions are available for this type of image. In Photoshop CS5, many adjustment layers and filters are not 32-bit compatible. The full repertoire of 32-bit functionality is only available in the Extended versions of Photoshop CS3, CS4, and CS5.

Image cropping and rotation is still possible, as is use of the Clone Stamp tool (e.g., for removing sensor dust specks). Although the Lens Correction filter isn’t available, several of the transformation functions used for correcting perspective distortion are.

In our example, we start by straightening the image using the tool, and then Image ▸ Image Rotation ▸ Arbitrary. We then correct the perspective distortion in the bridge’s towers using the free transform tool (Edit ▸ Free Transform). The Hue/Saturation command is also available to correct specific image areas selectively – e.g., by reducing the luminance of an otherwise flat-looking blue sky.

Figure 9-43 shows the range of adjustment layers that support 32-bit files. One of the most important functions is Exposure (see Photoshop Help for more information). However, if you make use of adjustment layers and want to retain the layer structure of your HDR image, you will have to save to PSD, PSB, or 32-bit TIFF, which will in turn destroy any compatibility with other HDRI programs. We therefore recommend that you merge your image into a single layer before exporting it for processing in non-Adobe programs.

Figure 9-43. Not all Photoshop corrections are available when working with 32-bit images.

However, the problem of visual inspection of the largely inadequate preview image remains. This is especially relevant in the case of corrections to color and hue. As previously mentioned, the 32-bit exposure slider in the bottom left corner of the preview window can help you gain a better overall impression of your image.

As already mentioned, the 32-bit exposure slider in the preview window (see Figure 9-44) can greatly improve the rendering of your preview image, helping you to get a better overall impression of your HDR image. The 32-bit exposure slider tool is very important to the decision-making process when correcting HDR images. But remember, the slider setting only affects the preview image, not the image file itself.

Figure 9-44. The 32-bit exposure slider in the Photoshop preview window provides a simple way to adjust the lighting in a preview image without actually changing the image file. This helps when planning exposure corrections.

It is nevertheless worth the effort to correct your images in 32-bit mode, as you will then have the largest possible reserves of quality and color depth. As a result, many of the potential negative side effects of your corrections will not be visible in the final image. Image alignment, for example, is best performed directly on HDR images (provided it hasn’t already been applied using Lightroom or Adobe Camera Raw). Adjustments made using ACR or Lightroom can be transferred to the other images in a sequence using the Synchronize command.^[136]

The same is true for cropping. Cropping can reduce the amount of data in a file, which in turn can speed up processing in general.

The next important step is tone mapping. This process (illustrated in Figure 9-45) reduces the increased dynamic range of your HDR image to that of a 16- or 8-bit LDR (Low Dynamic Range) image. The resulting LDR image can be reproduced more naturally and is supported by a much larger range of processing functions and filters. Printing is currently only possible for 8-bit images, or (with some limitations) 16-bit images.

Figure 9-45. Converting from HDR to LDR reduces the dynamic range of an image to 8 or 16 bits per color channel. The details of the algorithms used to do this remain the secret of the software companies that make them.

Figure 9-46. Photoshop CS3, CS4, and CS5 offer four different tone mapping methods.

The dialog shown in Figure 9-47, will appear when using Photoshop to convert an image from 32-bit to a 16-bit or 8-bit format. It can be accessed using Image ▸ Mode ▸ 16 Bits/Channel. Here, you can select your tone mapping method, and the results of your selection are displayed directly in the preview window. There are four tone mapping methods available in Photoshop CS3, CS4 and CS5: Exposure and Gamma, Highlight Compression, Equalize Histogram, and Local Adaptation. The functionality of the CS5 version of Local Adaptation has been greatly enhanced!

The best way to choose a method is to experiment and see which best suits your personal taste and your type of images. Not every method works equally well. The first three apply a global tone mapping which processes all pixels with the same tonal value, resulting in a new LDR image in which these pixels also have the same (possibly changed) value. You can usually achieve better results if the algorithm you use takes neighboring pixels into account. This type of method is described as localized tone mapping.

Photoshop offers the Local Adaptation tone mapping method for exactly this purpose (see Figure 9-47). This is the most versatile method. In addition to a number of sliders, you can use a gradation curve to fine-tune your adjustments. (The gradation curve that can be displayed in the other tone mapping dialogs can be viewed but not edited.) We use this method to tone map most of our images. We begin by moving the black point to the far left and the white point to the far right of the available tonal range using option B Toning Curve and Histogram.

If you click on the brightest and darkest points of the image using your mouse, Photoshop indicates the position of the equivalent luminance value on the gradation curve with a small circle icon, helping you to find suitable black and white points. However, the histogram can be misleading, especially in the highlight areas, which are sometimes hard to identify and thus easy to clip. Using the eyedropper tool together with the info window (Window ▸ Info) and the circle mark in the Local Adaptation window can often provide more accurate analysis of your image than a purely visual approach.

Figure 9-47. The CS5 version of the Local Adaptation method offers a full set of controls.

Once the black and white points have been set, you can start experimenting with the numerous sliders. We usually start by setting general brightness using the Exposure slider and then fine-tune the setting (if necessary) using the Gamma slider. We then brighten any dark shadows using the Shadow slider and darken highlights that are too bright using the Highlight slider.

The Detail slider adjusts microcontrast in a similar way to the ACR and Lightroom Clarity control.

During this process, Photoshop generates a brightness map which is never actually seen by the user and which works similarly to a soft-focus filter. The Radius controls the size of the area around each pixel taken into account while tone mapping; setting the value to “1” will only include the four immediately neighboring pixels.

The Strength value (called Threshold in CS4) controls the degree of blur, or, more accurately, how great the difference in tonal values between neighboring pixels should be in order for them to be grouped and equalized. In other words, the threshold value sets the maximum difference in brightness between neighboring zones. The threshold value here has nothing to do with the sharpness of the image. A higher value – strengthened by using a large radius – can cause halo effects to appear at high-contrast edges, as illustrated in Figure 9-49.

Unlike the standard gradation curve, the Local Adaptation tone mapping method does not process all pixels of a certain luminance uniformly. Instead, it couples the effect of the gradation curve with that of the luminance mask, making a certain amount of experimentation necessary to find the right levels for your particular image. As with all other complex processes, it pays to vary the slider values slowly, allowing Photoshop time to update the preview image between changes.

Figure 9-48. The Local Adaptation tone mapping method uses a virtual luminance mask.

Finally, we fine-tune the gradation curve, as illustrated in Figure 9-50. This is a process that requires a good deal of trial and error while you search for the place on the curve where the tonal value of a specific image area lies.

Normally, Photoshop smooths the curve. You can, however, create corners in the curve by creating a new point and activating the Corner option (A, Figure 9-50). This is not an effect you would want to apply to most normal gradation curves. As in the standard Photoshop Curves dialog, a curve point can be removed by selecting it and pressing delete.

Figure 9-49. A halo effect can appear at high-contrast edges if the threshold and radius values are set too high.

Figure 9-50. Our image after Local Adaptation tone mapping using the settings shown on the right. Unfortunately, the sky is still a little bland, so we will optimize it later in Photoshop.

Finally, we fine-tune the colors using the Vibrance control and (sometimes) the Saturation slider.

Figure 9-51 shows one of many possible versions of our processed HDR image. We used Photoshop to improve color rendering (using the Hue/Saturation command) and heightened contrast slightly before enhancing the details on the bridge structures using DOP EasyD Plus DetailResolver^[137]. We protected the sky from unwanted filter effects using a layer mask.

Figure 9-51. Our sharpened HDR image after post-processing with Photoshop

Instead of converting an HDR image to LDR using Photoshop’s tone mapping function, you may want to use a third party filter. HDRSoft’s Tone Mapping filter [77] is a good choice. This plug-in delivered better results than Photoshop for the example shown here. Its handling is very similar to Details Enhancer in Photomatix Pro, so we decided not to describe it in detail here.

That concludes our discussion of Photoshop-based HDR techniques. Because of its superior loading and aligning functionality, we often use Photoshop to generate HDR images that we later process using either Photomatix Pro or FDRTools.

Unfortunately, many of our attempts at tone mapping using Photoshop CS3, CS4, or CS5 turned out badly, even using CS5’s new, enhanced Local Adaptation method. When you are working with Curves, you cannot see which parts of the image are clipped as the clipping view (usually controlled by pressing the key) does not work. We do hope that Adobe improves things in the next version.

Our example uses four images of a sunrise at a mountain lake in Austria. Figure 9-54 shows the source images and their corresponding histograms. The photos were captured using a Nikon D200 and an exposure increment of one f-stop.

Figure 9-54. Our four 16-bit TIFF source images. The aperture was constant and the shutter speed was varied from shot to shot.

Clicking Generate HDR image in the program panel activates the image selection dialog. Our personal workflow involves selecting the images in Lightroom and exporting them to Photomatix using File ▸ Plug-in Extras ▸ Export to Photomatix.

In the next dialog (Figure 9-55 and Figure 9-56), we select our image processing options, including alignment type, ghosting removal, and gradation curve.

We usually use the By matching features option for aligning our source images. The settings for avoiding ghosting artifacts should be largely self-explanatory. Grass and leaves blowing in the wind and fast-moving clouds should be thought of as background movement. If your shot includes moving cars or people, you can use menu A (Figure 9-55) to set a higher level of detection. Higher levels will lead to longer processing times and don’t always improve the results.

With regard to the tone response curve (TRC) – here called the tone curve – we recommend the default setting, which generates a curve based on the color profile stored in your image.

Figure 9-55. Setting the HDRI generation options in Photomatix Pro

If you are using RAW source files, you can set the additional options shown in Figure 9-56. These relate mostly to white balance and color space. As most cameras select and embed the correct white balance (color temperature) setting into RAW files – at least when shooting outdoors in daylight, as was the case here – you should set the white balance to As Shot. Otherwise, you can select an appropriate color temperature from the list in the White Balance drop-down list. It is also possible to set white balance later during tone mapping – an option that gives better visual control. We usually set the ProPhoto RGB option for Color primaries because it offers the largest color space. We transform our image to a smaller color space later (after tone mapping), once we scale it for output. ProPhoto RGB is not a suitable color space for use with 8-bit images. 8-bit images (JPEGs, for example) are better processed using the Adobe RGB or sRGB color spaces.

Figure 9-56. Additional options available for processing RAW files

Clicking the Generate HDR button in the dialog shown in Figure 9-55 and Figure 9-56 generates an HDR image and displays it in a preview window (Figure 9-58). As with most HDR programs, the preview image is only provisionally tone mapped. The HDR Viewer window provides a better quality display, although it only displays a small detail of the entire image based on the last mapping process applied by the program (Figure 9-58). The detail at the current mouse location is shown in tone-mapped form.

We now save our new HDR image using File ▸ Save HDR as. In addition to the Radiance and OpenEXR formats, Photomatix Pro offers the 32-bit floating point TIFF output format (Figure 9-57). This is equivalent to the Photoshop 32 Bit (Float) format, but we generally use the Radiance format in order to preserve the greatest possible compatibility.

Figure 9-57. Photomatix Pro 3.1 HDR output formats

Figure 9-58. The first preview version of our HDR image. A better view can be seen in the detail displayed in the HDR Viewer window.

We like Photomatix Pro because of its great tone mapping functionality. Starting tone mapping either in the Workflow Shortcuts panel or in the Loupe panel produces a provisional tone-mapped version of your image with an optional magnified preview (Figure 9-62). The image already looks a lot better here. The preview size can be adjusted using menu A, and switch B toggles between the tone-mapped preview and the original HDR image. You can speed things up by limiting updates to the loupe detail using the option C rather than applying every slider movement to the entire preview image.

Again, the preview only shows provisional tone mapping, but this time the details are determined by the settings made in the mapping panel. Photomatix Pro offers two different tone mapping methods: Details Enhancer and Tone Compressor. Both have many more control options than Photoshop, making the Photomatix tone mapping process more versatile as well as more complex.

The Tone Compressor global tone mapping method is similar to Photoshop’s Highlight Compression process, but offers more settings (Figure 9-59). Here, we start by setting the black and white points. Although we can also use the current panel to adjust Color Temperature and Color Saturation, the actual compression and tonal value adjustment are controlled using the Tonal Range Compression and Contrast Adaptation sliders. The Contrast Adaptation slider adjusts the influence of the average brightness in relation to the intensity of the processed pixel. Moving the slider to the right tends to result in more pronounced colors. Moving the slider to the left tends to result in a more natural look.

Figure 9-59. Control panel for Tone Compressor tone mapping.

These settings alone produce a better result more quickly than we can achieve with Photoshop (even using Local Adaptation). Figure 9-61 shows our initial result, which could nevertheless do with some further optimization.

But, we are still viewing a preview. The actual image can be different in a number of respects. In order to view the actual results using either Photomatix Pro or Photoshop, it is necessary to convert our image to LDR (Low Dynamic Range). Start the conversion from HDR to LDR by clicking the Process button in the tone mapping panel (Figure 9-59).

Figure 9-60. Photomatix Pro’s LDR output formats

You can then save your image using File ▸ Save as (Figure 9-60). We always use TIFF 16-bit as our LDR export format.

Figure 9-61. Our image after tone mapping using the Photomatix Pro Tone Compressor method.

The real strength of Photomatix Pro lies in the Details Enhancer tone mapping option. This tone mapping method often gives a typical HDRI “grunge” look for many images, an example of which can be seen in the preview window in Figure 9-62.^[140] The illustration next to it shows the Details Enhancer tone mapping control panel. The Details Enhancer method applies a localized tone mapping similar to Photoshop’s Local Adaptation, also recognizable from the similarities between the controls offered by both processes.

We start by setting Strength to a moderate, mid-range value. The default value (70) is often a good starting point, but your results will be more natural-looking if you use values of 50 or less. We then need to make our basic settings for the black and white points, which we can fine-tune later if necessary – the higher the values we set, the more highlight and shadow detail will be clipped. A good working gamma value is 1.0. Higher values lead to higher- contrast images, while lower values produce duller results – which is not always a bad thing. Here too, you need to be patient while waiting for your preview image to refresh. It is all too easy to overcorrect your image if you try to work too fast. Be patient!

Figure 9-62. The preview of our first Details Enhancer tone mapping, showing the activated HDR Preview window. The look of the preview image results from using a high “Strength” value.

Figure 9-63. Details Enhancer panel

We have found that the following method is a good way to get used to the effects of the various sliders: working from top to bottom, set the slider to its extreme left-hand position and wait for the program to update the preview. Then set it to its extreme right-hand position, followed by a central setting. Now, continue to work with increasingly small slider movements to left or right until you find the value that appears best to you. There is no “perfect” setting, and you will always encounter situations in which small readjustments are necessary. Often, you will have to go back to a slider to do some more fine-tuning.

We use the Color settings tab to select a suitable color temperature (Figure 9-63) before we adjust shadow and highlight saturation. Slight increases liven up an image, but overadjusting saturation can quickly produce an artificial look.

Now we adjust micro-smoothing in the Miscellaneous settings panel (Figure 9-64). Micro-smoothing is similar to microcontrast. You can find more information on the subject of microcontrast (also known as “local contrast”) in Enhancing Local Contrast. A higher value for Micro-smoothing can help to liven up an image and make further contrast adjustment at post-processing unnecessary. Adjustments to microcontrast in Photomatix are applied to the entire image, whereas Photoshop allows you to apply them selectively using layer masks.

Figure 9-64. Miscellaneous settings panel

If you are using high Strength values, you may also need to increase the level of Micro-smoothing to keep the microcontrast effect under control.

We now move on to the Shadows Clipping setting. If shadow detail is less important to the overall look of your image, you can select a high value to produce a high-contrast image with dark shadows. The best settings for your image depend on the subject itself and your personal taste. Our example uses a value of 2 in order to retain the shadow details.

Increasing the Highlights Smoothness and Shadows Smoothness values can help avoid shadow and highlight clipping by adjusting the range of highlights or shadows that contrast enhancements are applied to. Moving the loupe over the corresponding sections of the image can help to determine the best values.

We can now go back to the main settings and once again adjust the Strength setting. The best value to use here is also largely a matter of taste. Many published HDR images are vilified for their exaggerated and sometimes unnatural look – a criticism that also applies to some of the images in this book. However, if you set a moderate Strength value (50 or less), you should be able to avoid the more pronounced effects that HDR processes can produce.

Another disadvantage of higher Strength values is increased image noise. We therefore recommend that you keep your Strength values low and that you shoot at low ISO speeds whenever possible. The Reduce noise option (available since version 3.1 – see Figure 9-55) does exactly what it says.

We now move on to the color saturation, contrast smoothing, and brightness settings, followed by a final gamma value adjustment.

We have described all these adjustments simply and sequentially, although in reality, many of them have reciprocal effects. You simply have to practice to find out which controls have which effects and how they fit in with your personal processing style.

Once you find a combination of settings that works well and might be suitable for other images, you can save it using the Presets drop-down menu (Figure 9-63). This will enable you to reuse the same values when processing similar images and save a good deal of time, even if the resulting image still requires a little fine-tuning.

Keep an eye on the histogram and check the individual color channels before producing your final image, as color clipping is not always visible in the simple luminance histogram display (Figure 9-65).

As before, the actual tone mapping is activated by clicking Process in the tone mapping control panel. The results are then displayed in the preview window and can differ significantly from the preview image.

Figure 9-65. Here, highlight clipping in the red channel is not visible in the combined luminance histogram.

The second tone-mapped interpretation of our HDR image is shown in Figure 9-67. This image has already been fine-tuned and slightly sharpened in Photoshop. The emphasis here is definitely on the word “interpretation”, because the program’s numerous settings give you a very high level of flexibility in your work. The real indicator of a good result ought to be what you saw or wanted to see while shooting. It is often necessary to process various versions of the same subject. For instance the dark sky and almost oppressive mood in this image is not suitable for every type of image.

Figure 9-66. Our final settings for the image shown in Figure 9-67

Figure 9-67. The second version of our image was created using Details Enhancer tone mapping method and the settings shown on the left. The resulting image was fine-tuned using Photoshop.

When you return from an extensive shoot with lots of new HDR material, you may find the prospect of processing each sequence individually time-consuming and daunting. You will nevertheless want to get an overall impression of the sequences which warrant more work – which is where batch processing can be really useful. Using a batch process, you can run all of your image sequences through the processor using standard settings and inspect the results. You can then easily select your best images for manual processing.

Photomatix has a useful, built-in batch processing function. The function does, however, rely on your source images being sorted according to specific criteria to work effectively. There are two ways ensure this:

If you shoot very different sequences, it can be an advantage to create several folder structures in order to keep your HDR and Exposure Blending sequences separate. You also need to decide whether your results should be saved in the same folder as the source images, or to their own, separate folder.

If you shoot RAW images but want your source files to be in a preprocessed TIFF format, you first need to process your images using a RAW converter. The exact process differs from converter to converter, so refer to your software manual for details.

Once you’ve organized your source material, you can start batch processing either by clicking the Batch Processing workflow shortcut button or by navigating to the Automate ▸ Batch Processing menu item. The dialog shown in Figure 9-74 is now displayed. The available batch processing settings are as follows:

Figure 9-74. The Batch Processing dialog contains nearly all the necessary settings.

Figure 9-75. This dialog is used to set a naming convention for your output files.

Figure 9-76. This image of a rainbow is a panorama made from two HDRI images, each of which consists of two separate source images.

Click Start to begin batch processing. Batch processes are run separately from the rest of the program, which means you can perform other, non-batch Photomatix operations while your batch process is in progress. The progress of the individual steps is detailed in the window shown in Figure 9-77.

Finally – or as soon as the first results are ready – you can view your finished LDR images and optimize tone mapping in the appropriate HDR images if necessary. This is easier to do if you save your HDR and LDR images in their own folder, separate from your source images.

Large batch processes can take time but usually run without problems. We often use the Photomatix Pro batch process to make a “first draft” of our HDR images before further processing them using other HDRI tools.

A simplified batch processing function for single files can be found in the Automate menu under Single File Conversion. This process tone maps single HDR images located in a specific source folder and saves them as LDR images in a target folder. Here, too, you can specify the type of tone mapping to apply and the output format.

Figure 9-77. The Photomatix Pro batch processing window details the individual steps as they are carried out.