Low light photography embraces many subjects and situations, from highly charged downtown city streets at night to silent evening landscapes.
In the last chapter we looked at what is for the most part a reportage approach to photography, suited to subjects that move, unpredictable moments that need quick reaction, and a way of working that is light on equipment and carries a percentage of failed shots.
But when the main subject remains still, or reasonably so, and you have more than a few seconds to get ready for a shot, the more reliable technique is to use a tripod, or some equivalent means of locking down the camera. This simple act opens the way for long exposures that can handle almost any light level, and some guarantee that the shot will be technically fine, meaning sharp, accurately exposed, and with noise kept under control. Some of the newest technical developments, especially in the processing stage with techniques like HDRI (High Dynamic Range imaging) can make a quite remarkable difference, enabling results that would have been inconceivable without digital technology.
There is, as we’ll see, rather more technology involved (or at least available to us) in locked-down photography than in handheld, and more of it concentrated in post-production. Nevertheless, while the process of shooting is certainly more reliable than handheld, it demands its own particular technical proficiency. There is more equipment to carry in the form of the tripod, and that alone establishes a different approach. In place of the physical attention that handheld photography demands, there is care in tripod management and protection from vibration—demands that have not, of course, changed throughout the history of photography. What is new is the need to plan ahead at the time of shooting, making exposures that will in many cases undergo considerable post-production.
INSIDE TO OUT
When shooting interiors with a view through to a bright exterior, one option is to lock the camera to a tripod and take several shots of varying exposures to composite later n editing software.
However proficient you become at handheld shooting, a solid, locked camera support is still the most secure technique for shooting.
It is essential for any lighting condition that puts the fastest shutter speed below the safe limit for the combination of lens focal length and hand steadiness. Tripods, in various forms and designs, are the basic mechanical support. Rigidity is paramount, though for ease of carrying around this often has to be compromised to some extent by the tripod’s portability. Increasingly, new materials, such as carbon fiber and basalt fiber are being used, which have a higher strength-to-weight ratio than traditional metal alloys, so are light, while remaining rigid, but are considerably more expensive.
The standard portable design has telescopic legs, usually in three sections, a rising center column for fine adjustments, and a fixed angle of spread. There are many variations as manufacturers compete for attention. A variable angle of spread has advantages in that you can accommodate different surfaces, but this usually involves an extra set of bars with sliding collars (these are standard for professional filming and video, where weight is not such an issue as in still photography). An alternative is a second fixed position (Gitzo use this extensively). A wide spread can also be useful for a lower camera position, and better stability in windy conditions.
QUICK SETUP TRIPOD
The Manfrotto 458B Neotec tripod is fast and easy to use, with an innovative Neotec rapid opening and closing mechanism—just pull each leg downwards to open and automatically lock it in any position, with no screws, knobs, or levers to tighten or loosen. To fold it back up again, press the release button and push the leg closed.
FLAT TRIPOD
The Cullman Universal Magic tripod has the ability to fold each leg flat for portability.
GITZO
The 3530LS Tripod is a simple and classic design from manufacturer Gitzo.
FOOT OR SPIKE
Tripod with changeable foot or spike.
MULTIPOD
The Cullmann 3080 Multipod can be braced both against the ground and walls or any other available upright at the same time.
ALUMINUM TRIPOD
Sturdy aluminum legs make this tripod firm but heavier than some other models.
TABLE-TOP TRIPOD
A small table-top tripod combines portability with rigidity, when you don’t need an eye-level viewpoint.
NOVOFLEX
This mini tripod design has legs that can be moved to different holes in the head for convenience.
CARBON-FIBER TRIPOD
The lightest tripods are built with carbon fiber, like this Manfrotto model, which is designed to take the strain of outdoor use while being as light to carry as possible.
CULLMAN MOSKITO
A portable mini-tripod which features a two-way pan head.
MARKET INNOVATION
As the behavior and tolerances of synthetic materials are better understood, more and more versatile products are emerging, like the Gorillapod, which has flexible legs that can be bent around any convenient support, or used in traditional mini-tripod form.
There is a surprisingly big difference in stability between a tripod used well and one set up inefficiently. Tripods obey the laws of mechanics, and the aim in setting up is for maximum stability.
This is largely a quest for rigidity, and there are a number of ways you can help. A tripod’s telescopic legs are more stable when compact than when extended, which means that a tripod at its most compact on a solid raised ledge (or other surface) will be steadier than the same tripod extended from the ground. Equally, when you extend a telescopic leg, extend the thicker sections before the thinner ones.
The surface on which the tripod is erected also matters. Outdoors, beware of sand and any loose conglomerate like pebbles. If in doubt, twist and press the feet down to make sure they are on as stable a surface as possible. Grass and earth, even though soft, can be good, provided that the feet can be pressed in—a spiked foot is an advantage. Remember that outdoors the tripod must also battle the wind, which is especially troublesome if it is gusty. Indoors, carpets are notoriously spongy and uncertain, and bare wooden floorboards can shift if trodden on, even some distance from the tripod.
The tripod is at its most stable when the head is exactly centered between the three feet which it sites above, and when the platform, immediately below the head, is level. Some models have a spirit level built into the platform. Try and level this rather than leveling only by means of the head. For the same reasons, try and ensure that the combination of camera and lens has its center of gravity over the center of the tripod. That’s why long lenses have threaded mounts.
You can test the stability by gently tapping the end of the lens. Look for any visible movement of the equipment, then repeat this tapping while looking through the viewfinder. As mentioned, you can improve the stability of the setup by lowering the center of gravity of the tripod plus camera. One additional way of doing this is to add a weight to the center; a few tripods have a hook at the base of the center column for this purpose. Hang anything you have handy, even your camera bag.
Tripod maintenance involves keeping the collars and other parts of the telescopic leg system clean, dry, free of abrading particles, and lubricated appropriately. Molybdenum disulfide and graphite are high-quality lubricants often used by manufacturers, but it also pays to find out exactly what the tripod manufacturer recommends. Also keep the joints at optimum tightness—not so tight that they are difficult to move, but without any play or sideways movement.
The shorter the legs, the more stable
Extend thicker leg sections first
Choose a solid surface
Level the tripod platform
Aim for the center of gravity of camera and lens over the center of the tripod
Shelter from wind
Add weight for stability
SUPER TELEPHOTO
This Nikon lens has a center of gravity much nearer the front, hence the support, which should be mounted onto a tripod rather than using the camera’s own mount.
TRIPOD BAG
Use a camera bag to add weight and stability to a tripod setup.
TRIPOD BAG
This LowePro camera bag includes a tripod holder making carrying of a full-sized tripod far easier.
MORE STABLE
LESS STABLE
There are two basic designs of tripod head, but many individual designs, and some crossovers. The most traditional is a pan-and-tilt head, with each axis of movement kept mechanically separate (roll, pitch, and yaw, if you like).
The entire assembly rotates at the base, and this can be locked, usually with a screw knob. Two locking arms each control one of the other movements. Rotation, of course, is 360 degrees, and the advantage of this tried and tested design is that small adjustments can be made in a single direction without upsetting the others. This is important if you need to level the camera precisely.
Ball-and-socket heads, normally referred to as ball heads, have become increasingly popular, partly because they can be switched between locked and completely free very quickly, and partly because they are so compact, adding very little size to the folded profile of the tripod. For situations that call for solid support at some times, but free movement at others, ball heads have a clear advantage. When slightly loosened, they are ideal for following subject movement, though heads designed for videographers have some advantages in this area.
OFF-CENTER BALL
An off-center ball head allows a different range of movements from the conventional center-ball design. Here it is combined with a separate pan control.
BALL HEADS
Increasingly popular for being compact and very quick to use, feature a single locking mechanism. The single disadvantage, which may not be important, is that releasing the lock means re-setting the camera completely, rather than being able to keep one or two axes aligned, as with a three-way head.
THREE-WAY HEAD
Made of light magnesium alloy, this typical three-way head offers three independent axes of motion, each with a locking handle/lever. The best brands use fluid cartridges for accurate movement with minimal effort.
QUICK RELEASE
Quick-release systems, like this one from Manfrotto, are an essential part of any quick-response tripod shooting. The base attaches to the tripod head, and individual plates to the camera(s).
There are many alternative solutions to tripods for securing a camera, adapted to different shooting situations and viewpoints.
SUCTION PAD
Activated by a U-shaped lever, grips on any smooth surface such as glass, marble, metal, or plastic, and takes a load up of up to 6 lbs (3kg).
WOODSCREW
A woodscrew with ¼-inch thread connection. Can be screwed into beams, fence posts, and tree stumps.
GROUND SPIKE
For pushing into any ground like sand, earth, pebbles, snow etc. It combines with a ball-and-socket head.
NOVOFEX CAR WINDOW ACCESSORY SET
Featuring an accessory plate with a screw thread for connection of a ball-and-socket head or other accessories. Rigidity is achieved by the additional support obtained from a monopod or the Novoflex Chestpod.
SLR SUPPORTED
Clamping to toughened glass can be surprisingly effective, even with quite heavy cameras.
CLAMP AND ROD SET
This versatile set handles all of the situations shown on this page, and more.
CLAMP AND ROD
Additional height can be achieved when a clamp is used in conjunction with a rod.
BAR MOUNT
The indent in the clamp means it can also be attached to a circular rod, like a handlebar.
Motion blur with the camera mounted on a tripod allows much greater control than when handheld, and gives two distinct possibilities.
One is subject motion blur with the camera motionless, so that the setting remains sharp. The other is smooth panning, and if you have time to set up a tripod, this assumes some kind of predictable, or anticipated subject. We looked at panning and tracking see here, and while either works well enough handheld, using a tripod offers two advantages. One is that it frees at least one hand to operate camera controls, the other is that it allows a steadier flow of movement. In either case, subjects that have distinct blocks of tone or color, and an outline that is easy to recognize, tend to convert well into a motion-blurred image. Using a tripod makes it possible to work with quite long exposures, such as one second or more, although the mirror-up position makes following a subject at these speeds a little unpredictable.
A word about the tripod head. Ideally, there should be enough friction to support the camera and lens without putting strain on your hands, and with enough free play in the direction of the pan to follow the target without drag effects. As most panning sequences involve horizontal movement, it usually helps to loosen the rotating axis fully, keeping the friction quite firm on the other movements. This depends on the kind of head. For instance, some ball-and-socket heads have an additional rotational movement; a good panning setup with this type is medium friction on the ball and loose on the rotation. If there is no additional rotational movement available then you will need the ball to be set looser.
One danger in tripod panning is that if the tripod itself is not level, the vertical and horizontal smear lines in the image from motion blur will become increasingly askew as the camera is rotated. If you have time to prepare the setup and know in advance the route that the moving subject will follow, check the moving view through the viewfinder with a quick practice run first. Another reason for checking out the path of the pan is that it may involve changes in focus and exposure. Most lenses are now autofocus, but for ones that are not, you will need to continually check and adjust the focus during the pan. Almost all modern SLRs have focusing features for just such situations as these, and can make some prediction of distance as the camera follows the subject. This is known as AI Servo on Canon cameras, for example, and can be quickly set via the camera buttons.
ENLIVENING WITHOUT DETRACTING
One practical use of motion blur with the tripod locked down is when the subject is inherently not particularly interesting. In this case, a photograph was needed of the entrance to this office, but to give it some extra interest, the shot was timed for the moment when a visitor was leaving. Setting a shutter speed of 1/4 sec made the person less recognizable, and blurred the figure to make it less prominent.
This sequence of images, with a figure clad in black walking repeatedly in front of an art installation, was shot at 1/3 sec. The shutter speed was determined in advance by experiment—sufficient to show that it is a walking figure, but as motion-blurred as possible to emphasize the ambiguity of the scene. The basic problem is that the movement of the legs in certain positions (most positions, in fact) makes the figure look one-legged. In cases such as these, it is important to shoot sufficient frames for a good choice. Here, the third frame of the four “reads” best.
As touched on earlier, “dark” or “fixed-pattern” noise is caused by imperfections and irregularities in the sensor, including, for example, uneven boosting between the individual readout amplifiers accompanying each photosite on a CMOS sensor.
It increases approximately in proportion to the exposure time, and at one or several minutes can be very noticeable. It is also enhanced by heat-generated electrons adding to the photosites, so you can expect more noise in warm weather than in cold. Because this kind of noise is related to the sensor and its circuitry and not to the image, it lends itself to fairly efficient correction. The key, however, is to record the exact pattern of the noise at the time of shooting. Some DSLRs have a menu option to do this automatically and make an immediate reduction of the noise, and the results are usually impressive. The main drawback is that it adds the same amount of time again as the original exposure, and if you’re allowing a few minutes for an exposure, this may not be acceptable.
ORIGINAL
This night shot, with a full moon rising behind pyramids in the Nubian desert, was exposed at 15 sec at an aperture of f/3.5 and a low ISO of 200. The clarity of the atmosphere accounts for the high level of brightness and thus relatively short exposure (see here). Long-exposure in-camera noise reduction was on. Nevertheless, while this combination of settings works well for an image as exposed, noise problems still appear when attempts are made to open detail from the shadow areas.
AS EXPOSED
Noise is well suppressed with the Raw image converted to TIFF at standard settings, without any adjustment to either the exposure or the brightness.
CORRECTED
This area of the image, with an area of moonlit sand visible, shows what happens when the Exposure slider in Photoshop ACR is raised by 2 stops. The result is serious noise, which would need to be addressed in post-production.
Perhaps surprisingly, the difference between noise-reduction on and off is quite subtle, even at the long exposure of one minute shown here in this side-by-side comparison. The top pair is a comparison between on (left) and off (right). To highlight the effect, a deliberate underexposure of two stops on the aperture was corrected in Raw conversion. The lower pair is the same, save that the exposure is correct. The difference between on and off is slightly more obvious on the upper pair. More importantly, adjustments in the Raw conversion (or with a TIFF by raising the gamma in Levels or Curves) has a more damaging effect than not choosing in-camera reduction.
2 F-STOPS UNDEREXPOSED, CORRECTED IN RAW
Noise reduction
on
Noise reduction
off
CORRECT APERTURE SETTING
Noise reduction
on
Noise reduction
off
While fixed pattern noise and, to an extent, readout noise are amenable to in-camera treatment, “random noise” is intractable.
By its nature it is completely unpredictable, and it is also difficult to gauge how much it contributes to the overall noise in any given situation. In the final image, the different sources of noise are overlaid on each other. However, there exists an effective, though time-consuming method of isolating and reducing “random noise” on a picture-by-picture basis. This is the image averaging technique, and calls for a few or several identically framed shots taken consecutively. Because it needs to be planned for, and drastically slows down shooting, it is not likely to become a standard technique. It can be performed only with the camera locked on a tripod, because the frames must be in pixel-perfect register.
The principle is that if the scene is completely still, without any subject movement, the only variable between identically exposed frames will be the random noise. Averaging the images removes this noise. In astrophotography this is a well-known technique, and it applies equally to regular low light photography. In practice, it means repeating the shot without any movements to the camera, taking special care with the tripod and shutter release. No special software is needed other than Photoshop, because the averaging is done with layers. Like Dark Frame Subtraction to reduce dark or fixed-pattern noise, it increases the signal-to-noise ratio of the image and so is non-destructive, unlike most software methods which blend noisy pixels into their local neighborhood. The noise between two frames is averaged, in effect reducing it by half in the combined image. This is more or less the equivalent of noise from half the ISO setting.
In Photoshop use the image stack Auto-Align features. First, copy and paste all the images into one layered image. An alternative is to load the layers with a script (File > Scripts > Load Files into Stack). Then select all the layers and run Auto-Align to put them into register. Convert the layers into a Smart Object and then choose a Stack Mode. The two useful modes for noise reduction are Median or Mean. Try both, but Median has the ability to replace pixels more drastically, rather than just reducing the strength of noisy pixels. Moreover, if there has been subject movement between frames, such as branches moving in wind, or a passer-by, Median can remove them. In any case, it is worth experimenting with both to see which is the more effective for a particular image.
ENLARGEMENTS SHOW NOISE
The original, shot in sequence at ISO 1600.
An alternative way to combine the multiple images is to use Photomatix, the popular HDR software. Photomatix has an Average option, which applies a median filter to a group of images.
In Photoshop, the same combining procedure can be achieved. It takes more steps, but is efficient. First copy all the frames into one image, as separate layers. Once the layers are merged into a Smart Object it’s a simple matter of clicking Layers > Smart Objects > Stack Mode > Median.
Original
Photoshop
Selecting all the layers, Edit > Auto-Align Layers is applied, using Reposition Only.
Once aligned, the layer stack is converted into a Smart Object. This is achieved by highlighting all the layers and clicking Layer > Smart Objects > Convert to Smart Object.
Result
The high contrast caused by pooling of light is typical of many low light scenes, from a restaurant interior with spotlighting, to a city view at night.
In the days of film this meant either bringing in photographic lighting to open up the shadows and balance the contrast, or simply accepting the high-contrast result with all its accompanying loss of detail. Now, however, this problem can be solved, particularly with a tripod. For subjects in which there is little or no movement, or at least where any movement is not a significant part of the image, the solution involves shooting the same scene at a number of different exposures, with no camera movement between frames.
As we’ll see over the next few pages, there are a number of ways of combining a range of exposures, but the principle is always the same—to take the best detail from each individual exposure. So, with a high-contrast scene, the exposure that best deals with the middle range will likely overexpose highlights and underexpose shadows. Once the highlights have blown, at 255 on the histogram, the sensor photosites have essentially been flooded, and there is no detail captured at all. Correspondingly, serious underexposure renders deep shadows black (level 0), and while dark shadow areas may retain some detail that can be pulled up during processing by using Levels, Curves or a local-contrast process such as Photoshop’s Shadows/Highlights feature, this will always be at the cost of increased noise. However, if in addition to the standard exposure that captures midtone detail, you shoot a longer exposure to capture shadows and a shorter exposure to capture highlights, you have effectively increased the amount of visual information you have to work with.
Another way of looking at this is that by bracketing exposures of the same scene, you are extending the dynamic range of the camera’s sensor. Admittedly, this is not all in one frame, and the next steps on the computer can be time-consuming and complex, but what is important is that you have captured much more detail than would otherwise have been possible. The bottom line is that if you have the time to do this, a range of exposures is a great back-up. You have lost nothing but a few megabytes of storage space, and gained a valuable archive of visual data. The following pages detail the current ways of handling this data, including the newest, HDRI, but we can also expect future imaging software to improve. The key lies in the capture.
The procedure is straightforward, but there are ways of making it more efficient and reliable. Clearly, the idea is not to move the camera at all between frames, and this makes a cable or remote release valuable, as well as good tripod use (making sure that it is firmly sited) and anything to reduce mirror slap, such as locking the mirror up. However, absolute precision to the pixel is not necessary. Desirable, certainly, but recent imaging software advances have made automatic alignment of images highly efficient to within several-to-many pixels.
TRADITIONAL
The best of the traditionally exposed images has blown highlights around the lights and central window.
HDR IMAGE
The result of merging several differently exposed frames shows more detail in areas that were completely blown in the example opposite.
One important alignment issue, however, is that you should vary the exposures by altering the shutter speed, not the aperture. Aperture affects depth of field, which can change the content of the image between frames, particularly as in low light you are likely not to be stopping down fully. There are two important considerations in working out how many frames and at what settings. The first is that the exposure gap between frames is short enough so that the software you use later for blending can handle them, yet not so short that you clog up the memory card unnecessarily and take too long in shooting. As a rule of thumb, certainly for HDRI, 2 stops works well in most cases. One stop is safer, but usually redundant. Time and storage space are important, and many cameras have a small frame buffer, which can mean that you have to wait for frames to be processed by the camera before you can continuing shooting.
The second key consideration is to set the right limits for each end of the range. You need the shortest exposure to capture all significant highlight detail, and for this the best guide is the camera’s on-screen flashing warning for highlight clipping. Nevertheless, you probably don’t need every tonal detail of naked lamps. Treat these as you would specular highlights, and let them blow. This highlight clipping warning makes the shortest exposure the most convenient place to begin the sequence. From this, increase the exposures regularly between frames. Continue until the last exposure captures all the shadow detail that you want. Be warned that at night, because of the eye’s adaptation and because the camera’s LCD screen is likely to be much brighter than your surroundings, it’s easy and normal to think that the exposures are brighter than they really are. As a precaution, and until you get used to doing this, continue at least one step beyond what you judge to be sufficient by eye. In addition, check the histogram display on the camera. For the last frame, capturing shadow detail, there should be a distinct peak on the right side. Failure to capture full shadow detail is the most common problem in shooting a range of exposures, and the penalty is noise.
First find (by measurement or trial and error) the shortest exposure which shows no highlight clipping. The most convenient indication is the flashing highlight warning. The shutter speed needed is the setting just shorter than this.
The first frame is the shortest shutter speed without clipping.
2 stops more.
2 stops more.
The next frame is shot at 2 stops more exposure than this, using the shutter speed control only.
The next is 2 stops more again.
2 stops more.
The final exposure is the one that shows the darkest shadows at the midpoint in the histogram.
The least technical, most personal method of blending different exposures in a layer stack is to use the Eraser brush on an upper layer, so as to reveal better exposed detail underneath.
In principle, nothing could be simpler, and it’s worth saying that while Photoshop experts might treat such a non-procedural approach with some disdain, for photographic realism this is in no way an inferior technique. Indeed, with some types of image it can be more efficient too. Typically, the kinds of images that respond well to manual brushwork are those where the difference in exposure between the layers is confined to distinct spatial areas of the image. For instance, in a shot through a doorway or window from, say, a dark interior to a bright exterior, it may be fairly straightforward to select this hard-edged frame and then erase this part of an upper, light, image layer.
Brush techniques tend to be idiosyncratic, in that most people simply develop their own particular ways. It is less valuable to try and categorize all of this than to show specific workthroughs. The examples here illustrate just some ways of doing this.
SOURCE IMAGES
The original three source images, spaced 2 /-stops apart.
The two darker images are pasted in as layers over the lightest, so that the layers are ordered darkest at the top, lightest at the bottom.
The middle layers are erased to leave just the exterior view.
Next, the top layer is made visible, and erased to leave all but the sunlit portions of the exterior.
The layer stack after the two upper images have been selectively erased.
The final result.
Consider a low light scene in which you have plenty of static fine detail that calls for a slow exposure that will be relatively noise free, but at the same time there is some movement that you would also like to capture.
The dilemma is whether to choose a low ISO setting and put up with motion blur in parts of the photograph, or a high ISO setting which will capture the movement, but increase the noise. Remember from what we’ve seen of the different causes of noise that dark or fixed pattern noise is much easier to reduce than readout noise at a high ISO.
Well, there is a way of having the best of both worlds, although it takes time, requires a tripod, and works only in certain circumstances. The principle is to make two separate exposures, one at a low ISO and longer shutter speed, and the other with a high ISO and short shutter speed. The two can then be combined later, choosing only the subject movement from the noisier high-ISO image. In effect, this drastically reduces the area over which you need to apply software noise reduction.
The example here illustrates the process and what it is capable of. The first image was handheld at 1/20 sec and f/4.2 at ISO 3200. With a tripod and, using the LCD screen view of the image to help align the new shot, I took the same view at ISO 100 and at 4 sec. The post-production involved running a noise-reducing filter over the figure of the old woman, combining this and the noise-free version in register as layers, and erasing the “noisy” background from around the figure. Reproduced small-scale as most of these pictures are here on these pages, the benefits are not so obvious, but practically this meant that the image could be printed acceptably up to exhibition size.
THE ORIGINAL
The original, handheld shot, ISO 3200 as shot without noise correction.
SIMPLE CORRECTION
The same, with noise correction applied using Noise Ninja (adjusted for best effect on the figure only).
TRIPOD SHOT
The second shot, taken at ISO 100, tripod-mounted.
The first image, processed in Noise Ninja for noise reduction, is pasted as a layer over the second, cleaner image. At full scale to begin with, the surroundings are erased from the top layer using a large, soft brush.
Next, the eraser brush is reduced in size, and the erasing continued at magnification.
COMPARISON
A detailed comparison of the key area of the image, where the figure and background are blended. From left to right: the first high-ISO noisy image, the same image processed for noise reduction using Noise Ninja, the noise-reduced version combined with the second, noise-free image.
The final result.
Currently the best-known dedicated software for blending a range of exposures is Photomatix, a cross-platform program that also handles HDR generation and tone mapping.
Although both blending and hand-tone-mapping have a similar purpose in compressing a high dynamic range captured in different exposures, the techniques are quite different. Of the two, blending is easier to understand, and the several Photomatix options follow the same basic principle as the highlight-and-shadow blending method described here. They are, however, more refined and automated.
Compare the description here with that of HDR and tone mapping on the following pages. Both methods use the same source files. Blending, which is called Exposure Blending in Photomatix terminology, selects highlight details from the less exposed images and shadows from the more fully exposed images. These are then combined in such a way that there is a smooth, unnoticeable transition across the midtones. The algorithms used make use of weighted averages, and in one of the five procedures offered this can be shifted by the user toward darker or lighter.
ORIGINALS
The three original source images, spaced 2 f stops apart
The algorithms for each are different, and as the results always need to be judged subjectively, as to how pleasing or otherwise they are to you, it is difficult to predict which to choose. Accordingly, Photomatix offers a batch-processing option, and recommends in case of any uncertainty using this to produce a range of results. You can then choose which looks best to you.
Average – This method combines images using the same averaging formula throughout the image, regardless of whether the computed pixels are in a dark or bright area. For this reason it cannot give the dynamic range extension possible with the other blending methods, but it has a natural effect and also has the effect of reducing noise.
Fuse exposures with H&S – Auto can handle more than two images, and is designed for a “natural looking” result. As a consequence, it errs on the side of flatness, with low contrast, but this can be improved later with any of the previously mentioned Photoshop contrast-enhancing techniques.
Fuse exposures with H&S – Adjust This gives you the choice of weighting the result toward the brighter or darker end of the range of source images, and also adjusting the radius (as this is a local-contrast enhancement method). A high radius produces a sharper-looking result, at the expense of a longer processing time, but runs the risk of halo artifacts around sharp changes in brightness.
Fuse exposures with H&S – Intensive This is intended to deal with a range of source images covering a very high dynamic range. As with Adjust there are a number of adjustable settings available, including Strength, Color Saturation, and Radius. With a high dynamic range, HDR and tone mapping is likely to be a
BATCH PROCESSING
The Batch Processing window in Photomatix. For this exercise, all the options have been selected. This is also a useful technique when there is uncertainty over which of the several blending methods will be best.
Average
Adjust
Auto
Intensive (strength 0)
Intensive (strength 10)
GLOBAL OPERATOR
For comparison, an HDR file generated and tone mapped using the global operator (Tone Compressor).
LOCAL OPERATOR
Also for comparison, an HDR file generated and tone mapped using the local operator (Details Enhancer) at its default settings.
ADJUSTED SETTINGS
The HDR tone mapped in Details Enhancer with adjusted settings. The Smoothing preset option has been set to Low and the Color Saturation increased creating the surreal saturated look common to many HDR images.
When I first wrote about High Dynamic Range Imaging in 2005, it was used more in CGI (Computer Generated Imagery) for controlling lighting effects than in photography.
Within a couple of years it was being used experimentally by a wide range of photographers eager to explore its ability to create hitherto impossible imagery. This is still a relatively new technology and we can expect more progress in techniques and tools. Nevertheless, it is tailor-made for several types of low light photography, including interiors with bright views out and bright city lighting. Both of these situations have huge dynamic ranges. I touched briefly on dynamic range here, but here it’s important to distinguish between low and high. These are relative concepts, and for photography, it is the practical issue that decides. Essentially, if the dynamic range of the scene can be captured in a single exposure, with the possible exception of specular highlights, this is low.
Technically, therefore, 8 bits per channel is a low dynamic range format. 12-bit and 14-bit, which you can capture using Raw with a DSLR, cover considerably more, but unfortunately most of the extra range is crowded into the highlights, where it is not really needed in photography, and little in the shadows, where it is. All camera sensors, therefore, work in a low dynamic range. In terms of scene lighting, this is usually adequate as long as there are no light sources in the frame, and as long as you do not need to pull out detail from visually deep shadows. Low lighting situations, however, often do feature lights in shot. An interior lit with only one desk lamp can cover 10 stops from close to the lamp to a dark shadow under the desk, while the lamp itself can extend this by 4 or 5 stops.
CHURCH INTERIOR
Another classic HDR situation is the interior of a large church, or cathedral. HDR and tone mapping is able to open up deep shadow areas in the vaulting, while also rendering the much brighter stained glass windows accurately.
High dynamic range formats typically have 32 bits per channel, which alone expands the range hugely. However, the normal formats and color spaces, as well as the performance of monitor displays, use non-linear gamma encoding, and an important effect of this is that more of the bits are devoted to highlight information than to the shadow areas. In fact, the more bits per channel, the less efficiently the image is encoded digitally. To overcome this, HDR formats use their 32 bits per channel not as discrete allocations of pixel values, but as decimal points. Floating point encoding assigns a precise value to every pixel, rather than forcing it into the nearest pre-determined level. The result is that an HDR file format can record a virtually infinite dynamic range.
LIGHT SOURCE IN VIEW
HDR and tone mapping easily meets the challenge of images in which the major light source is in view, as in this modern tea-ceremony room in Tokyo. Note also the holding of highlight detail in the spotlit wall-hanging.
An HDR format capable of recording these high ranges is one thing, but capturing is another. The technique involves making a series of exposures at different shutter speeds. In practice, 2 stops is the most convenient gap between exposures, and in order to avoid changing the structure of the image, it is better to alter the shutter speed than the aperture. A basic and effective way of shooting is to begin with an exposure that avoids clipped highlights, using trial and error and the camera’s on-screen highlight warning. This is the shortest exposure of the series. Then lengthen the exposure time by 2 stops—4 times the exposure, in other words—and take the second exposure. Keep doing this until the darkest shadows in the scene appear fully exposed; this looks like gross over-exposure on the LCD screen.
There are a number of software options for combining a series such as this into a single HDR image. The result is a 32-bits-per-channel floating-point image that contains all the data from all of the exposures. It is, however, unviewable because nearly all monitors are 8-bit, and even the few with extended range are a modest 14-bit. The second part of the operation, therefore, involves compressing all of this data back down to 8 bits or 16 bits. There are a number of ways of doing this, but without any predictable best method. The essential difficulty is that the detailed execution of compressing a huge range into a small one depends on the nature of the image and on your interpretation of it. An HDR image has the potential to mimic human vision in its ability to take in a high dynamic range, but the eye and brain work quite differently. The subject is complex, but in perception the eyes scan in rapid, jerking movements known as saccades, and these multiple “recordings” are rapidly assembled in the brain into a coherent view. There is no way in principle of mimicking this in one flat image. The various stages of making an HDR image are covered in the following pages.
Typically, the process of creating an HDR image begins with capture, which we dealt here.
The ease of creating an HDR file depends very much on this stage, because while the generation is largely automated, the software expects certain things from the captured sequence. It needs to access the exposure information, which is carried in the EXIF header, the exposures need to be appropriately spaced across the range, the several frames need to be in reasonably close alignment so they can be matched, and for the same reason there should not be too much movement of objects within the scene. If all these conditions cannot be met there are ways of working around them, but with more difficulty.
An HDR generator is software that collates all the different exposure information from a sequence of exposures and then creates a single HDR image file. There are a number of generators available, some of them free, and while there are technical differences, for normal photography there is little to choose in performance. The most widely used are those supplied with Photoshop and Photomatix. In use, all are relatively quick. Once the range of captured images has been selected, the generator looks for the exposure data in the EXIF headers.
As both Photoshop and Photomatix can read Raw files, if you are shooting Raw there is no point whatsoever in first opening and adjusting the images with a Raw converter. All the information can be accessed directly by the HDR generator.
Once the 32-bit-per-channel HDR file has been created, it appears on screen as an extremely contrasty and sometimes dark image. A regular 8-bit monitor is no more able to display the full range of the image than an 8-bit file format is able to record all this exposure information. The viewing experience of an HDR file is predictably unsatisfactory, but to be practical there are few reasons for wanting to be able to look at the image at this stage, other than to check that nothing has gone seriously wrong, such as an image from another sequence wrongly added.
The final step in HDR generation is to save the image file in one of several formats, although many people choose to bypass this and go straight to tone mapping, in which case the HDR file is converted and lost. The two most commonly used formats are Radiance and OpenEXR. There are important differences between them, but more from the perspective of CGI and motion-picture special effects, where accuracy across the entire range and file size (because of the huge number of frames in, say, an animated movie) matters. From a photographic perspective any of the available choices in the Save As dialog will do. If you plan to archive the HDR file, though, you may want to consider one of the more compact formats, such as OpenEXR.
ROOFTOPS
Generating the HDR file is only the first step in creating the sort of HDR image that we are increasingly familiar with. The HDR file has to undergo a process known as tone mapping before it can be viewed or printed.
The final stage in HDR Imaging is to create a viewable 16-bit or 8-bit image from the 32-bit HDR file.
This is neither predictable nor particularly easy to manage, although there are several software methods. Much depends on the original scene, how big a dynamic range it has, and the way you shot it, but for various reasons simply compressing it to fit a 256-level scale is rarely satisfactory. If there are very bright highlights, for example, a simple proportionate compression will result in an overall dark image with insufficient attention paid to the midtones. Because the many tones in the HDR file have to be assigned somehow to the much more restricted low dynamic range (LDR) file, the procedure is called tone mapping. The software that performs this operation is known as a tone mapping operator, sometimes referred to by the shorthand TMO. The several tone mapping operators available are the result of various ingenious algorithms that attempt to give a result that is a compromise between retaining tonal detail across the scale and looking realistic. This last, realism, is the biggest problem in tone mapping, because it represents a cross over between the subjective and the psychology of perception.
All of this means that in practice there is a considerable amount of trial and error, and you may often want to try out more than one operator to see which looks best for you in which situation. There are two classes of tone mapping operator (TMO); global and local. A global operator works on the entire range of tonal values in much the same way, as does, for example, a tone curve on any image (think of normal Curves or Levels operations in Photoshop). Global operators have the advantage of giving results that look photographically normal, therefore realistic, but they have the disadvantage of usually sacrificing detail and contrast in some areas of the image. A common example of a global operator is a pair of sliders controlling gamma and exposure. Playing with both will usually produce an image that is acceptable but often flat.
The second class of TMO is the local operator. These algorithms, which vary considerably in the principles on which they work, adjust the tonal value of each pixel according to its neighbors. This makes it possible to adjust local contrast, and potentially this is a very powerful way of mapping tonal values. Accordingly, one of the key adjustments here is the radius—the distance around each pixel that is searched. If you are already familiar and comfortable with other radius-based procedures, particularly unsharp masking and Shadows/Highlights in Photoshop, you will become accustomed to these local operators more easily. The advantage of a local operator is that it can handle even wide dynamic ranges with surprising efficiency. The disadvantages are that the results can often look false and over-detailed, there is a danger of halos around distinct areas (such as the sky close to a window frame or a horizon), the results are difficult to predict and the user settings are non-intuitive. Local operators need considerable assistance from various controls.
In both cases, global and local, it is normal for the image to need further work. Getting the tone mapping right in one go is difficult, often impossible. Local operators in particular are likely to look rather flat. For this reason it is essential to produce a 16-bit image from the HDR 32-bit file. After further adjustment, which would otherwise lose data and result in a spiky, toothcomb-like histogram, you can then reduce it to a final 8-bit image. The most usual post-procedures for tweaking a 16-bit file are Curves or Levels, Shadows/Highlights, and even the mild local contrast enhancement of USM used with a large radius and small amount.
FULL-SIZE HDR
HDR tone mapping makes it possible to shoot in very contrasty lighting, such as in a bar and restaurant without special lighting, and still achieve an attractive and controlled result.
Color saturation is another issue. Basically, the more extreme the gradient of tonal mapping, the more saturated the final color is likely to be, and this often results in certain colors appearing far too rich. Some local operators have a color saturation control, but in any case, you can expect to need to perform some adjustment on the output 16-bit image. The Hue/Saturation sliders are an obvious choice, but a more targeted procedure is the Replace Color tool in Photoshop’s Image > Adjustment menu.
Tone mapping is, above all, very specific. HDR software designed for photographers began with Photomatix, was followed by Photoshop, and now includes a few more, notably FDRTools, EasyHDR, and Nik HDR Efex Pro. There will undoubtedly be more to come. Each works in a particular way, different from the others.
PROCESSED COMPARISON
In the processed comparison, the best original frame (LDR, low dynamic range) is on the left; in the center is the HDR global tone mapped version and on the right the locally tone mapped version, both in Photomatix. Global tone mapping does a satisfactory and realistic job of combing the exposures, but comes nowhere near the local tone mapping operator, which holds detail in each tonal neighborhood of pixels, from highlight to shadow.
SEPARATED BY TWO F-STOPS
Enlargements of a detail—one of the spotlit dishes from the restaurant image (see here)—shows the original sequence of 3 frames, each separated by the standard 2 f-stops.
CHROMATIC ABERRATION
Lateral chromatic aberration is exaggerated strongly in the process of generating and tone mapping an HDR. This small detail from the corner of an image, uncorrected and corrected, gives an idea of the extent of this problem.
Photomatix is the pioneer HDR application for photography (as has already been noted, HDR has other uses, notably in computer-generated imaging), and in the current crop of available software is noted for the large number of user controls that it offers. One result of this is that it enables a wide range of possible treatments for any image, including some highly non-realistic ones. Photomatix offers a basic choice between Tone Compressor (its global operator) and Details Enhancer (its local operator).
As with other HDR software, the global operator sticks close to “photographic” realism, but is much less effective in tone mapping, so that with a wide range of exposures in the original sequence, the local operator—Details Enhancer—is the usual choice. The default setting gives a reasonable and realistic result under most circumstances, and it is best to begin with this. Photomatix offers a high level of control over all settings and more recent versions offer useful presets as starting points.
LOADING BRACKETED PHOTOS
Using Photomatix as a standalone program, begin the process by going to File > Load Bracketed Photos, and navigate to the image files.
PREPROCESS OPTIONS
Once you’ve clicked OK, the preprocess options include alignment of images based on content, always a useful precaution although it adds a little extra processing time, ghosting removal (unnecessary if there is no subject movement), noise and chromatic aberration reduction, and White Balance.
PREVIEWING
Once generated, the HDR image is previewed. Go to Edit > Tone Mapping/Fusion to start processing the image. Photomatix, as well as providing a good level of control for either Tone Mapping or Exposure Fusion processes, also offers a number of presets that appear as small thumbnails in a window running under the preview image. When you’re happy with your settings press Process.
The Photoshop tone mapping procedure offers four ways of tone mapping, of which three are of little use. Exposure/Gamma is a straightforward pairing of an exposure slider (for brightness) and a gamma slider (for contrast combined with brightness), which offers very basic global tone mapping. Highlight Compression applies a custom tonal curve to reduce highlight contrast drastically so as to brighten the rest of the image and restore contrast. Equalize Histogram, applies a different custom tonal curve to even out peaks in the histogram. All of these are global operators. The most useful is Local Adaptation. In 16-bit mode this offers a good level of control over four key elements of the image, Edge Glow, Tone and Detail, Color, and Curve. There is also a limited number of presets to choose from.
PROCESSING
To begin the HDR process in Photoshop, go to File > Automate > Merge to HDR Pro. In the ensuing Merge to HDR Pro dialog box select the Raw files.
GENERATION
The HDR Photoshop generation process is automatic, and uses the layer system.
ADJUSTMENTS
Before the HDR is finally created, a preview window appears, showing the exposure settings calculated, and offering user adjustments to key elements of the image. There is also a selection of presets, although these are of limited use.
Nik software’s HDR Efex Pro is one of the more recent examples of HDR software. It has an attractive, user-friendly interface, with controls that will look familiar to anyone used to using image-editing software.
AS A PLUG-IN
When used as a Lightroom plug-in, simply select the various exposures of the image in the Library mode, and go to File > Export with Presets and select HDR Efex Pro.
Nik’s HDR program offers localized adjustment in the form of Control Points. These allow you to make exposure, saturation, and contrasts adjustments to specific areas of the image.
EASE OF USE
Nik HDR Efex Pro is easy to use, offers a good level of adjustment and control, and features a variety of presets (visible down the left-hand side of the dialog box) that can be selected from a list of popular photographic genres.
This piece of cross-platform software is notable for integrating the creation of HDR files from the source images with tone mapping. In other software this is a separate operation from tone mapping, with FDR Tools it is combined with the tone mapping in the sense that at any point during tone mapping you can return to the HDR creation and make a number of adjustments.
These include choosing which tonal ranges are used from which source images, tweaking the choice of images to use for ghosting removal, or even opting to drop one or more of the images. Although the interface is not as intuitive as other HDR software, FDR Tools is extremely powerful and versatile.
In the HDRI Creation panel it’s possible to make tonal adjustments to the individual source files before finalizing the HDR process. These adjustments are previewed so you can see how they are likely to affect the outcome.
FDR Tools provides three tone mapping options. Simplex and Receptor are global operators, while Compressor is the local option. Many of the controls as well as being operated by sliders, can also be adjusted via curves, allowing you to apply levels of adjustment to specific tones. These provide a very high degree of control over the final image.
Ghosting is the term given to superimposed, displaced parts of an image, quite common in any multiple-exposure technique when parts of the subject move.
If you are shooting a scene with passers-by, for example, then it is inevitable that they will be in different positions in each of the frames. Moreover, as this is low light photography, some or all of the frames are likely to show a degree of motion blurring.
Many HDR generators, however, have some provision for attempting to remove ghosting. This works typically by detecting the variance in the image—the areas where the content differs—and then by giving priority to one frame of the sequence. This works most successfully when there are very few frames and movement in just one or two frames, as in the example here. Varied movement across a large number of frames, and in which the movement overlaps, can cause problems. It also helps when the movement is in an area with a similar tonal range, as movement across a strong luminosity edge, such as a skyline, can also cause failure. FDRTools notably allows adjustment of exactly how the frames are prioritized for ghosting removal, a useful option.
Finally, consider using the high-low ISO blending technique described here. In this case you can expect tonal differences, but as the purpose is to select just moving subjects, such as people, this should not normally cause too many problems. When shooting, take one or a few frames at a sufficiently high ISO to allow a reasonably short shutter speed. The frame should in any case be in register with the HDR image. Paste on top after the HDR image has been tone mapped to 16-bit, check that it is in register by clicking this upper layer on and off quickly, and if any repositioning is necessary, do this now with the arrow keys. Next, erase around the objects you want to keep to replace their ghosted versions in the tone mapped HDR image, and adjust the tone and color to match, using Curves or Levels. Then you will need to flatten the image.
Illustrated on this spread is a simple situation for ghosting, with just two differently exposed frames, and so hardly HDR, but it illustrates the technique well. The two images were used to generate an HDR file in Photomatix, with ghost removal activated. For comparison, the result without ghost removal simply appears as a double exposure.
FIRST FRAME
The first frame, exposed for the sky and highlights on the pagoda.
SECOND FRAME
Less than a minute later, the second frame, exposed for the shadow areas in the foreground, where the three people are walking.
DETAIL COMPARISON
In close-up, the area with the variance. On the top, without ghosting removal during HDR generation; on the bottom, with.
GHOSTING REMOVAL
Choosing the ghosting removal option during the HDR generation step prioritizes the second frame for the foreground.
STANDARD COMBINATION
Combining the two images as a 32-bit-per-channel HDR image in Photomatix and then tone mapping. The blending of the two images without the option of removing ghosting has the predictable result of a 50% blend across the movement.