Shooting extra frames to guarantee sharpness is good practice when the lighting conditions are pushing the limits on camera shake, motion blur and accurate focusing (see here).
What then usually happens is that the sequence is examined carefully and at magnification to find the best shot—technically and creatively. The choice may be obvious, but sometimes you may find that two or three frames each have some advantage. The focus may be sharper in one but the composition better in another, or there may be unwanted subject motion blur in the sharpest version. In this case, there is another solution, though it may not appeal to everyone. This involves taking the best of one and adding to another.
The principle is straightforward—layering the images in register, then erasing the unwanted parts of one. As long as the frames are matched for their content, color, brightness and contrast, this gives you the freedom to have the best of both.
Under the circumstances illustrated here, where we are essentially dealing with a rapid burst of shooting, matching color, brightness, and contrast is not normally an issue. Typically, load the frames together into the Raw converter (you’ll notice that I’m assuming throughout this book that everything is shot in Raw), Select All, and adjust as necessary. Even if there were exposure differences between frames, simply toggle back and forth from one frame to another and adjust until they are visually matched. More demanding is getting the frames in register. The manual method is tedious: copy one frame onto another as a separate layer, set the Opacity of the upper layer to somewhere around 60 percent so that you can see both images but one is more prominent, zoom in to 50 percent or 100 percent, and use the arrow keys to align them as well as possible. When they seem to be in register, set the Opacity of the upper layer back to 100 percent and check the register by toggling the visibility of the upper layer off and on repeatedly. Registration becomes more difficult if there has been rotation between the images, because to adjust this (in Edit > Transform > Rotate) you need to see the entire image, making exact alignment harder to judge.
This operation has been made much more feasible with Photoshop’s Auto-Align Layers function. First lock the layer you want to use as a reference (if it is the Background image it will automatically be locked), then select both or all of the layers that you want to align, including the reference layer, then go to Edit > Auto-Align Layers. The Reposition Only option should normally be sufficient if there was little camera movement between frames, but the Auto option will automatically make whatever geometric transformations are needed. Note that because this automatic alignment is based on the software’s analysis of content, it may make compromises that you don’t like. For example, if there has been a parallax shift of foreground details due to slight movement in the camera position, the alignment process, which is of course based on image content, may base its calculation on foreground or background.
Finish the alignment if necessary by using any of the tools in Edit > Free Transform, and/ or making tonal adjustments to even out any exposure or color differences. Once the images are aligned and matched for tone and color, use the Eraser for a direct replacement of less-sharp areas. An alternative to the Eraser is to mask the layers which is completely nondestructive, but the undo tool is always available in ordinary editing and is quicker and easier to access, and creates smaller files.
While these adjustments are not absolutely critical in this case, this sequence of three shots (taken within the period of a single minute) allows a final composite that both eliminates the streaking of moving vehicles, and shows a denser mass of pedestrians. The workflow goes as follows: first compare the images and work out a plan of what to keep and what to eliminate, then cut-and-paste them into a layer stack, align them and finally erase the unwanted elements.
CHOOSING THE IMAGES
Examining the three images side by side, I could quickly devise a plan. In the center is the reference image, in which most of the traffic is stationary, bar three moving vehicles. The left-hand frame has much the same traffic and can be used to replace the vehicle streaks, while the right-hand frame has generally more crowded sidewalks, and the building on the right has a more easily readable blue-tinted video screen.
AUTO-ALIGN
Begin the Auto-align process by clicking File > Automate > Photomerge and selecting Auto.
PHOTOSHOP LAYERS
The three images are added to the same image as layers.
ERASING
Using the Erase tool or, if you are familiar with them, the Masking options, parts of the upper layer are removed to reveal the similar image from below. Here, for example, the blurred van is removed.
RAW PROCESSING
All the images are processed at once in ACR.
From a sequence of nearly identical frames, all shot at 1/15 second with an optically stabilized lens, there were a few shake-free and also a choice of the position and actions of the priest, who was cleaning the altar of this large and busy temple in southern China. Combining the sharpest frame, free of camera shake, with the preferred gesture of the man, allowed the best of both worlds.
FIRST CHOICE OF SUBJECT
The preferred priest image.
FIRST CHOICE BACKGROUND
The preferred background.
AUTO-ALIGN
Using the Auto-align function in Photoshop, the two images are registered. As the two separate layers show, there is an offset.
TEMPORARY OPACITY
In order to erase only those areas with underlying matched background, the upper layer is set to 50%.
ERASING UPPER LAYER
The upper layer is erased selectively, here made obvious by making the lower layer invisible for purposes of demonstration. Note that because the two pictures were not framed identically, the very top is left untouched, although it has some camera shake; this is judged to be unimportant because it is away from the picture’s center of interest.
FINAL RESULT
The layers are finally collapsed and the image cropped for the final combined result.
As anyone who has had the experience of aiming at a target knows, whether with a rifle or in archery, holding steady on a fixed target is by no means easy, both mechanically and psychologically.
Indeed, the longer you try and hold, the more likely involuntary movements are (discussed here). Introduce movement, however, and things become more fluid and steady, due to inertia. The same applies to handheld photography and the time-honored technique of panning: swinging the camera from one side to another to follow a moving subject can produce surprisingly sharp results at slow shutter speeds. The principle is that you make use of inertia (mass × velocity) to overwhelm smaller jiggles. When combined with vibration-reduction technology in the lens (see here), this can be a surprisingly effective technique in keeping at least some part of the image sharp.
Panning is the action of moving the camera to hold steady on a moving subject, but the reverse situation is also possible—moving the camera from a moving vehicle to hold onto a stationary subject. The detail effects are different, but the principle is the same. There is always an amount of uncertainty in doing this, but it is quite often worth trying, just to see what turns out. The instant feedback from digital lets you judge immediately whether to save or delete, and whether to continue with different settings.
There are some perceptual factors at work, too. Just as the range between sharp and blurred focus is greater with a fast lens used to isolate a detail at full aperture to help the focused subject to stand out, so the motion smearing from a panning shot helps to accentuate the part of the subject held in crisp detail.
Increasing the time that the shutter is open, and particularly over half a second, makes the outcome quite uncertain, not only due to prolonged subject movement, but because camera shake makes a bigger contribution. The effect then merges with the subject of the following pages, intentional motion blur.
HIGHLAND STAG
Using a 400mm lens from a standing position, a Scottish Highland stag captured at 1/60 sec and ISO 400.
TRACKING VEHICLES
A different kind of tracking, also from a moving vehicle, in this case a boat in the Gulf of Thailand at dusk. The shutter speed of 1/3 sec with a focal length of 200mm is far too long for a truly sharp handheld shot, even with vibration reduction activated, but the combined motion of the two boats, plus the steadying effect of a slow pan, give an interesting result that achieves just-acceptable sharpness in the man at the tiller.
MIGRATING BEAST
White-eared Kob beginning their annual migration across the Nile valley in southern Sudan, photographed with a 180mm lens from a height of about 100 feet, through the open door of a light aircraft. With the aircraft traveling at around 100 mph, the antelope running at around 30 mph, and a reverse pan with the camera (against the direction of movement), this is a complicated tracking shot that manages to hold most of the animals sharp at a shutter speed of 1/35 sec, wide open at f/2.8.
Motion blur is not all bad. Despite all the efforts that we normally make to avoid it and correct it, which we’ve just examined in more than usual depth, there are occasions when it can actually make the image.
Most photography has some documentary quality to it in the sense that it is showing a lucid view of something—a place, an event, a person—but photographs also have graphic qualities just as do paintings. The streaking and smearing effect of significant blur reduces the information content, but it can make the image more interesting, even intriguing. Sometimes photographs work better if they don’t reveal their content all at once.
In particular, motion blur does best what its name suggests. It gets across the sense of movement, and it does so in a graphically dynamic way. Moving objects photographed with a shutter speed sufficiently short to freeze them against their background have a curiously static quality at odds with the event. There may not be any clues in the picture to show that a car, for instance, was actually in motion. It could have been parked, and even if we know that it was moving in real life (a car in the fast lane of a highway is unlikely to be stationary), the quality of speed is often missing in the photograph. This does not necessarily matter at all, as we’re accustomed to the way photographs look, but it does open the opportunity for using a slow shutter speed to enhance the sensation of speed.
The key to all of this is occasional use. Using motion blur frequently as a graphic device simply becomes a mannerism—it is more effective when it appears as a surprise. Shooting handheld means introducing movement in to the image by the way you move the camera. This adds to any movement in the subject for an effect that can be complex and difficult to predict, which is one of its attractions. There are a number of examples from the history of photography, and one of the most notable were the several “motion” series shot by Ernst Haas in the 1950s, on subjects as diverse as bullfighting, rodeos, and motor racing. The swirling and streaked smears of color give an impressionistic effect. The results are always specific to the situation, so this technique calls for experimentation. Two key variables are the shutter speed and the kind of movement you introduce by shifting, shaking, jerking, or rotating the camera.
REVEALING MOTION
A Sufi adherent in Omdurman, Sudan, performs an ecstatic whirling dance. Fading daylight gave the choice of a sharp image at a high ISO, or the much more evocative and dynamic result chosen here by using a shutter speed of 1/30 sec.
VERY LONG SHUTTER
Panning with a long lens (400mm EFL) and a very slow speed, the aim of this shot was to convey, just, the top-hatted attire of brokers in a City of London street —a deliberately impressionistic treatment that was felt to be stronger than a literal and sharp image.
GETTING NOWHERE FAST
A planned shot of a hotel gym, where a slow speed of 1/4 sec at f/9 was needed in order to give some sense of action to what otherwise could have been a lifeless image.
ON SCENE
As part of a feature on traffic accidents in Bangkok, this overturned pickup, its driver already taken to hospital, was photographed with a deliberate jerkiness and at 1/3 sec (at ISO 800) for a sense of movement.
Using the Raw format for shooting is now the professional standard in almost all kinds of photography (news, sports, and when fast delivery overrides post-production are the notable exceptions to this rule).
Its two most outstanding benefits are first that it preserves the capture data separate from whatever settings were applied at the time of shooting, so allowing color temperature, for example, to be chosen later. Second, it can record a wider dynamic range than other formats.
Given the typically major shifts of color from one scene to another in low light photography, and the fact that they are often unpredictable, the settings-change features alone make the Raw format almost essential. Most digital photographers are by now familiar with the concept of Raw, but to summarize briefly, these are proprietary image formats devised by the camera manufacturers, each one different, with the intention of receiving post-camera processing in specialist software tools. Saving images in Raw rather than going straight to JPEG in-camera gives you access to the original capture data. All Raw converter software programs—and there are now many on the market—provide image-editing tools to extract the maximum usable information from the data, and convert it to a useful format such as TIFF, JPEG, or DNG (the open-source Digital Negative format promoted by Adobe for archiving Raw files). Between the converters there are variations in the number of features, how the controls work, and the way they read and interpret the data (much of which is undocumented, being part of the camera manufacturers’ code). Nevertheless, all do a good job.
High contrast is also a feature of many low light scenes that contain artificial lighting. This is partly because of the pooling of light from localized sources, and partly because the light sources themselves, such as desk lamps or street lights, appear in shot. High contrast scenes have a higher dynamic range than normal, and importantly, higher than can be rendered by a standard 8-bit image. Bit-depth is directly related to the dynamic range that a format can cover and, in the case of professional and serious amateur cameras, the sensor works in a higher bit-depth, typically 12-bits, but some at 10-bits and others at 14-bits. In other words, the sensor is capable of recording a greater range from highlights to shadows than most monitor screens and all printing devices can display, and at some point this has to be compressed or tone mapped (see here). In non-Raw shooting, the camera does this, but a Raw converter and computer are a more powerful combination, have more sophisticated features, and above all, allow you to decide on the final result.
As Raw converters develop and improve, they allow more and more of the normal workflow to be conducted within them. The raw converter in Lightroom, for example, now contains an adjustment brush and spot healing tools. This makes for a more convenient, one-stop way of working, so that under many circumstances you can make a final save directly from the Raw converter window without needing to open the image again and use other pixel-editing software.
ADOBE LIGHTROOM DEFAULTS
For reference, this is the image as shot in Raw, opened as default in Lightroom with no changes. A large area is marked by the clipping warning (red pixels).
RECOVERY AND FILL
Of the various ways of approaching the translation from Raw to JPEG, this method focuses on maximum recovery of highlight clipping and an appropriate fill. The first step here, as the sun is obviously well overexposed, is to increase Recovery until the clipping warning disappears.
RESTORING SHADOWS
As the Recovery darkened the overall brightness of the image, Fill Light is increased, which has the effect of restoring the shadows.
RESULT
The combined Recovery and Fill Light adjustments have predictably lowered global contrast in the image below, so the Contrast slider is increased, creating this result.
TRADITIONAL CURVES
A more traditional approach is to go straight to the Curve window in the converter. In Lightroom, four sliders below the curve target different tonal ranges, and here, the Darks and Shadows sliders have been used to raise the values of the second quarter of the range, while the Lights and Highlights reduce exposure in the sky.
VIBRANCE
Vibrance is Lightroom’s control for raising saturation without incurring clipping in any of the color channels.
DXO
For comparison, the Raw conversion in DxO, making use of the before-and-after window. All of the same issues are addressed here as in Lightroom, but with different algorithms and in a different order.
In principle, the full range of a high-contrast image captured in a 12-bit or 14-bit Raw file can be compressed satisfactorily into an 8-bit viewable file using a Raw converter.
The range of tools, including exposure, brightness, contrast, and curves adjustment, allow a great deal of control. However, in many Raw converters most of these are global operators—tools that work on the entire range of tones and colors in the image—whereas many low light situations have, as we’ve seen, sharply divided areas of brightness. The problem is often that of maintaining local contrast while holding or reducing global contrast. Photoshop’s Recovery and Fill Light sliders do an excellent job of local control at different ends of the tonal scale, but in an image such as this, it may still be difficult to achieve a normal, natural-looking contrast in both the darker and lighter areas.
One way of overcoming this problem is to make two separate conversions of the Raw image and combine them later. In one version, ignore the shadow areas and concentrate on making the optimum adjustments for just the highlights and the high middle tones (what Photoshop calls Highlights and Lights). In the second, do the reverse: concentrate only on the shadows and low middle tones (in Photoshop Shadows and Darks). This method borrows some techniques from multiple exposure blending—which normally requires a tripod (see here)—but is well suited to single-shot handheld photography.
In normal Raw processing, and indeed in all normal Curves and Levels adjustments, the operations are biased toward the mid-tones. The most common of all adjustments to a histogram is a shift with its center in the middle. Points on the histogram or curve at a distance from the center are correspondingly weaker, with the shadows and highlights being affected least of all. Even if the center point of a shift is taken from lower or higher in the curve, the same principle applies. Global operations, which is what these are, affect the parts of the image that are usually the most important—the mid-tones—and have the great advantage of being easy to understand, and realistic in effect. However, if as in the image here, the important areas are in the shadows and/or highlights, it may be better to treat them separately.
Note that local operators such as Photoshop’s Recovery and Fill Light adjustments are designed to tackle just this problem, and are very effective. They do, however, have one potential drawback, which is that because they work on local areas of pixels, they are not completely intuitive and can create unrealistic effects, particularly if aggressively used. The example above right shows what can happen.
Having made the two versions of the same image, the next step is to combine them. Here, contrary to what you might expect, there is a strong argument for doing this by hand with a large eraser brush, rather than procedurally. There are two reasons for this. First the circumstances for choosing double-processed Raw are usually when you have spatially well-defined areas, which are therefore easier to choose with brushwork. Secondly, procedural blending methods do not discriminate between which areas of the image are adjusted.
Finally, more recent versions of Aperture and Lightroom (and Photoshop as they share the same Raw processing algorithms) allow for an increasing level of localized adjustment. This began with a simple but effective spot-removal tool, but has now advanced to brushes that allow you to select specific areas or elements of an image and apply a whole variety of adjustments—tonal, color, contrast, saturation, and so on. Furthermore any amendments you make are non-destructive, living the original data left untouched. Using the adjustment or quick brush in this way obviously saves time and speeds the workflow.
BEST SINGLE RAW
The problem in the image above is that there is interesting detail in the straw raincapes hanging close to the camera, yet even using Recovery and Fill in the Raw converter fails to do justice to this while holding the bright reflections in the tin roof awning beyond.
RESULT
The combined Recovery and Fill Light changes have predictably lowered global contrast in the image below, so the Contrast slider is increased, creating this result.
LIGHT
DARK
DUAL CONVERSION
Instead, two Raw conversions are made, one concentrating only on the raincapes, the other only on the scene in the courtyard beyond. The darker version is pasted over the lighter, and partially erased with several strokes of a brush set to 40% opacity. This manual blending works because the area that needs treatment is well-defined and localized.
FINAL MERGED RESULT
The result, equivalent (though by very different means) to traditional darkroom print-dodging.
ADJUSTMENT BRUSH
Lightroom’s adjustment brush provides a good selection of possible amendments, which can be applied to specific areas of the image. Brush controls, too, are flexible, allowing you to set brush size, feather size, flow, and density.
MASK
Having selected the brush type and size, paint over the area you want to adjust. As well as making the adjustment, the brush also creates a mask overlay indicating the area affected and the density.
FINAL RESULT
Here’s the adjustment clearly showing the lighter rain capes. At any point you can go back and fine-tune the adjustment. As well as increasing exposure, I also increased Clarity, Saturation, and Contrast to bring out their texture.
High Dynamic Range Imaging (HDRI) is the technique in which a series of different exposures, all of the same subject, are combined into a single, special file format that can encompass a much wider range of tones than a normal JPEG.
We deal with this more thoroughly in the next chapter, because it demands that all the image frames are in perfect register. This naturally makes it a technique for tripod shooting. However, recent improvements in automatic image alignment and in DSLR camera design now make it feasible without a tripod.
In a way it is inconvenient to split up the subject between chapters, but unavoidable given that this is a handheld possibility, albeit specialized. For now, here is a brief summary of HDRI. Shooting a sequence of identically composed images that vary only in the shutter speed allows you to cover a potentially huge range of brightnesses. In this form it is, of course, useless, as it is spread across several images. However, there are now a number of special image file formats that can contain mathematically all the information in a single file. Known as 32-bit floating-point formats (see here for a more detailed explanation), they can encode an almost infinite range of brightnesses, from deep shadow to details of light sources. As ordinary camera sensors are incapable of capturing such a wide range in one go, the only practical possibility is to take a number of frames and then combine them into a single HDR file. This file must then be processed further in order to be able to produce a “normal” image file that can be viewed on a display monitor or in the form of a print, and this procedure, known as tone mapping, is discussed here.
As each frame should ideally be in register for this to work, HDRI has been very much linked to tripod shooting. However, software that can generate HDR files from a sequence, including Photomatix Pro, Nik HDR Efex Pro, and Photoshop now includes an automatic alignment feature, in which the images are analyzed for content and then shifted slightly in relation to each other as necessary. There are limits as to how well the auto-alignment can succeed, but there are no real difficulties with up to several pixels misalignment.
This brings us to the second part of the equation. One alternative to fixing the camera on a tripod in order to shoot the sequence is to shoot very quickly, in a continuous burst. This minimizes the risk of moving the camera between frames. Some DSLRs have an auto-bracketing function, and this can be turned into a useful way of shooting an HDR sequence. In the example here, the exposure bracketing was set up in advance with a Nikon camera. The highest step possible between frames is 1 f-stop, and the maximum number of frames is nine. While 2 f-stops is generally considered the most useful step between frames for generating an HDR file, 1 f-stop across nine frames gives a total range of 8 f-stops, which is perfectly adequate for most situations. Setting the camera to Continuous high-speed mode allows a burst of nine frames in as little as two seconds (this depends on the lighting conditions, as slower exposures add to the overall time). There should be little problem in holding the camera on the target for this short time, particularly if you are using a vibration-reduction lens. The main limit is the slower shutter speeds, with the risk of camera shake.
HANDHELD HDR
The sequence above begins with the best possible single frame, followed by the series of shots fed into the tone mapping tool. On the right is the aligned and mapped result.
Some HDR software, notably Photomatix Pro and HDR Efex Pro, although not the current Photoshop, will import single Raw images for tone mapping.
This is not, of course, true HDR, in that the range in a single Raw comes nowhere near the dynamic range that can be captured in a sequence of different exposures. Nevertheless, what makes this technique worth considering in low light is that the tone mapping procedures are sophisticated and powerful, and can often recover more from the shadows and highlights, as well as emphasizing local detail, than a normal Raw converter.
The example worked through here is processed in Photomatix Pro, which converts a Raw file into a pseudo-HDR file. In fact, the dynamic range of a Raw file depends only partly on the bit depth (typically 12-bits or 14-bits per channel). The file format at these bit depths is capable of storing a considerable dynamic range, but in practice this is much less, and is limited mainly by the sensor design, and specifically by the amount of noise present in the shadows. We’ll see more of noise in the following pages, but the practical limit to the dynamic range of a single capture is the “noise floor,” as it is called. This is easy to demonstrate if you open any image containing significant shadows in a Raw converter. As we saw before, modern Raw converters have ways of recovering information at both ends of the scale, but if you simply raise the exposure slider at a particular level you will simply be revealing noise in the shadows, not real detail. This is the “noise floor” and sets the lower limit. The upper limit, naturally, is set by full exposure, which clips highlights to pure, featureless white.
As always, each image is different and specific, and has its own needs, and this is particularly so with low light in which the shooting is usually pushing the limits. This photograph, shot at night on board a Thai fishing boat certainly presents a technical challenge. The only light for the captain at the wheel was a single bare lamp, but I liked the starkness of the lighting and the extreme play of light and shadow. Nevertheless, there was some doubt as to whether it would be possible to make a printable image.
JPEG ORIGINAL
The accompanying JPEG shows the image as shot—not very promising. I always shoot Raw plus a basic JPEG, the latter for quick preview and mailing purposes.
Using Photomatix Pro as a Lightroom plug-in is convenient for workflow. Exporting a single Raw file from Lightroom to Photomatix delivers this message.
Photomatix Pro now comes with a series of preset options as well as keeping all the individual controls. This is the result using default Tone Compressor.
Even with the many controls possible, no single tone mapping is ideal, and I decide to go one step further with this technique, and make more than one tone mapped version, with the intention of combining them later. In this second version, the default Details Enhancer settings lighten the shadows.
Noise in the smooth shadow areas on the left of the image continues to be a problem. For safety, I make a third version, this one aimed entirely at giving the least noise in these areas, ignoring the effect on the rest of the image. Shadows Smoothing is increased to the maximum under the Miscellaneous tools.
The three versions are combined in Photoshop, using layers, and hand-brushing with the Eraser. The main preferred version is on top, with the lighter open-shadow version underneath. The interior of the right window, and part of the center window behind the man’s head are brushed through to reveal more detail. This done, these two layers are merged (Merge Visible), and the noisy areas in the shadows, particularly at left, are brushed away to reveal the less noisy version underneath. Finally, the layers are flattened.
The man’s face appears too blown out and pale. The image is opened in Lightroom and using the Adjustment Brush in Color mode a warmer glow is applied to his face.
Lightroom’s new noise algorithm offers a great level of control. I can significantly reduce the noise in the darker shadow regions, while holding detail in the man’s face—by far the most important element in the photo.
The final result is a major improvement on any normal method of Raw processing.