The ends of the day, life indoors and the entire range of night-time activities offer a rich and large source of subjects for photography, now more accessible than ever before. And it is digital photography, with its many continuing advances that has made it more accessible. What we are talking about here is extending the shooting range into situations that, because of insufficient light, could not be handled as freely as film photographers would have liked.
Photography has always had technical limits, but then photographers have always learned to accommodate and work within them. In fact, they are not always thought of at the time as limits, and it is only when these are about to be broken, or have just been broken, by technological improvements that we can see just how restricting they have been. The quantity of light and the sensitivity of the recording medium has always been one of the most basic limitations. In the early days of photography, it restricted shooting to bright daylight, and even then the exposure times were not short. As film emulsions improved in sensitivity, and as lenses became faster and cameras smaller, it became possible, but only just, to shoot in artificial light. The first photojournalist was, arguably, Eric Salomon, who made use of the 1924-designed Ermanox 858. This camera, designed by Ludwig Bertele, had a maximum aperture of f/1.8 and shutter speeds up to 1/1200 second, making it possible for Salomon to capture his celebrated “candid” pictures of society life in Berlin. Developments were rapid, and Leica soon announced a camera with similar capabilities that accepted the more convenient 35mm film rather than rigid plates.
Digital photography and its recent changes have ushered in a new era in low light photography by extending the range of time and situations that can reasonably be captured, making it possible to use a camera under conditions that would have been unthinkable without resorting to invasive techniques such artificial setups and flash. Sensor sensitivity has improved hugely in recent years, and this, combined with new in-camera and post-production processing, has reduced the collateral damage to images from noise—a major preoccupation in this kind of shooting, and in this book. In this first chapter, we’ll look at how the digital process—from sensor to in-camera processing, to post-processing—handles light when it comes in quantities significantly less than during a normal day.
ARTIFICIAL LIGHTING
Shooting interiors can often present white-balance challenges due to artificial lighting, as well as those of low light levels.
As we’ll see throughout this book, low light photography pushes limits. It pushes the limits of technique, demanding constant attention to steadying the camera, subject movement, and changing camera settings to suit the conditions.
It also pushes the technical limits, beginning with the sensor and lens. The special conditions found in many low light situations tend to highlight deficiencies of the sensor. Notable among these are blooming and noise. In effect, these operate at opposite ends of the brightness range; blooming is flare surrounding overexposed highlights, while digital noise is at its most apparent in the shadows, in longer exposures, or at higher sensitivities.
Blooming is the leakage from a photosite on the sensor that has reached full capacity (complete white) into adjacent photosites. High-quality sensors show less of this than cheaper ones, but some flaring is inevitable in the typical low light situation where there are bright lights against dark backgrounds. The city night view here is typical. Nevertheless, we are so accustomed to seeing flare in images that it is not necessarily a problem. In most photographs, the flare from the lens is greater than that from sensor blooming.
Another defect noticeable in isolated highlights—or along strong luminosity edges—is chromatic aberration. There are two kinds, axial and lateral, and in modern lenses it is the lateral aberration that tends to be the culprit. The role played in this by sensor blooming (see here) is disputed, and is probably tied to a different aberration known as “purple fringing.” Lateral chromatic aberration shows itself as two opposed colors, usually red-cyan or blue-yellow, and is increasingly common because of wide-range zooms, which are difficult for lens manufacturers to correct across the range. Also, every photographer is now a few clicks away from a magnified view of every image, so these things are simply noticed more. There are several solutions depending on the software. Most typical is a manual operation, moving a slider, and this works by expanding one channel relative to another. There is also defringing software that works on the edges themselves. A more automated solution is to use a pre-calculated lens module that corrects for the aberration in known lens-camera combinations, such as DxO Optics Pro and Lightroom.
Another detail from the Tokyo night shot, highly magnified, shows the effects of blooming, or flare, when highlights are overexposed. Sensor blooming occurs when photosites fill up and the charge leaks to adjacent wells. The left image was shot at 1/3 sec and the center one at 2 sec. In the 2 sec exposure light is spilling over from the edges of the neon display. The right-hand image is the 1/3 sec exposure processed for an increase of nearly 3 stops to bring it to approximately the same brightness as the center. This creates a very noisy image with exaggerated fringing, but blooming is not noticeable to the same degree.
PURPLE FRINGING
This detail of a light source from a night shot shows purple fringing, believed to be caused by sensor blooming or from the microlens, or both. It differs from normal chromatic aberration in that it is a single color in one direction.
CHROMATIC ABERRATION
A night view of Shibuya, Tokyo, shows lateral chromatic aberration from a wide-angle zoom.
Noise is one of the special problems that arrived with digital photography, although it is less of an issue now. It appears as a random pattern of pixels, usually bright and multicolored, superimposed on the image. The comparison that is frequently made with grain in film is far too generous, as grain structure in an emulsion can contribute a gritty texture that is possible to like. Nobody to my knowledge has ever made a case for digital noise being aesthetically pleasing. In terms of appearance (rather than cause), there is luminance noise (the random pattern of pixels varies from dark to light), chrominance noise (the pixels vary in hue), “dead” pixels (bright dots) and JPEG artifacts, if you shoot JPEGs (blocks of 8 x 8 pixels can be prominent as these are used in the compression). Noise is a particular issue for low light photography because whether you increase the ISO for handheld shooting, or increase the exposure time for tripod-mounted shooting, noise will be created.
A simple comparison of noise at different camera settings. Noise is always most apparent in smooth areas that are dark, but above black. Practically, the major difference in noise comes from changing the ISO sensitivity. This sequence runs from an ISO setting of 100 to 6400. A point to note is that in terms of appearance, these noise effects are exponential—meaning they set in quite late on the scale, but at the high end of ISO choices quickly become objectionable.
The second most common kind of noise is dark noise from long exposures, and here, at a one-minute exposure, the noise is less of an issue with this sensor and this camera than the random noise from high ISO. There is a barely noticeable difference between the camera’s dark-frame subtraction process switched on and switched off. The temperature is critical, however. This sequence was shot at a reasonable room temperature of 21ºC, but this kind of noise tends to double every 6ºC to 8ºC.
LONG-EXPOSURE COMPARISON
Left is 60 sec NR on, while right is 60 sec NR off.
ISO 100
ISO 400
ISO 800
ISO 3200
ISO 6400
We will look at noise and noise reduction in detail in Chapters 2 and 3, but in practical terms it is essential to be thoroughly familiar with the appearance of noise as created by your camera in images that you typically shoot.
Few people would disagree that noise is best avoided, but the amount of effort you go to in order to control it should depend on how important it is to you. Long-exposure noise can generally be taken care of quite well in the camera by selecting the noise-reduction option in the menu, but the immediate cause of most noise is setting a high-ISO sensitivity. You may well find that for certain kinds of image, you can tolerate a higher ISO setting than you imagined. As an important first step, make some tests at different ISO settings in a range of low light situations, and examine them side by side. At 100 percent magnification, you can expect to see increasing noise with higher-ISO settings most obviously in smooth shadow areas. Decide for yourself at what setting it becomes objectionable. As noted on the previous pages, you need to find an acceptable balance between the usefulness of a high-ISO setting and the degree of noise.
THE CASE FOR LOWER KEY
For shots taken in low light to convey that impression, the key of the finally processed image needs to be lower than normal. This is entirely a matter of taste, but here, the darker of the two versions is more faithful to the original impression—something that only the photographer can decide. The histogram is always a good check—the bulk of the tones centered is a normal key; shifted to the left is low key.
There are several variables when it comes to making this decision. One is the medium that you would normally use for displaying your images. If this is a print, then you should make tests using your preferred paper type. The appearance of noise will be different between a 100 percent screen view and a print. A related variable is the size at which the images will be used. A fully magnified print will obviously show more noise than a 640 x 480 pixel screen image posted on a website. A third variable is the kind of image. A reportage shot with visible noise is likely to be tolerated by more people than would be a still life or landscape. Yet another variable is knowing how much you can expect to, or are willing to, reduce the noise later with software.
Ultimately, when shooting, you should be armed with enough information and judgment so that you can confidently select the appropriate ISO setting. You may, for instance, see hardly any difference in noise between shooting at ISO 100 and ISO 200, in which case the extra shooting speed carries no penalty. The ISO setting is one of several ways of achieving a certain shutter speed, and the others include supporting the camera, sacrificing depth of field by opening up the aperture, and switching to a faster lens. Each carries some disadvantage, even if minor, so consider the ISO selection in this context. Maximizing shooting efficiency may mean altering the ISO frequently, and some cameras allow it to be changed by means of a dial, which is faster than going to the menu.
As well as technical matters of exposure, such as holding highlights and avoiding shadow noise, which are all the more critical with artificial light sources, you need to decide whether a scene should look dim rather than fully bright—in other words, the key of the shot. With a night image, this is perfectly expected—it would look strange otherwise—but there are many less obvious situations where there is room for interpretation. The example I’ve chosen here is deliberately not a clear-cut case: a Hakka communal dwelling in China, where the atmosphere inside the circular three-storey structure, open to the sky, was distinctly subdued.
In one important sense, low light photography is not just another situation, or just another themed area of photography. It carries with it a challenge; that in order to work in these lighting conditions you are forfeiting predictability.
The lighting conditions themselves are never quite sufficient to allow the perfect camera settings, so not only are you close to reaching the shooting limits, but you will always have to make technical compromises. If you sacrifice shutter speed, you run the risk of motion blur or camera blur. You can trade this against a number of other things, including aperture, noise from a higher ISO setting, lens focal length, viewpoint, the moment of shooting and more besides. The list of choices go on, and are always specific to the scene in front of you. The point I’m laboring here, because I think it worth a couple of pages, is that low light photography has built into it the idea of pushing the envelope. The reward, as just mentioned, is being able to work in, and capture, an area of life, that was up until recently fairly restricted to photography. The price is a higher level of concentration on technical matters, and a higher failure rate, which simply takes some getting used to.
Photography as a whole is premised on there being enough light to work easily, and the norm is full daylight. There’s no precise definition for this, but it goes from around 8 a.m. to 5 p.m. Not surprisingly, cameras and their sensors are designed for trouble-free operation in these conditions, meaning that the sensor can operate at its lowest, cleanest sensitivity for highest image quality and yet still allow a choice of shutter speed and aperture. These in turn give you flexibility in capturing motion and in deciding on the appropriate depth of field. In order to shoot successfully in the more restricted conditions of low light, you lose any flexibility. On the contrary, none of the settings are likely to be what you would really like.
WIDE ANGLE, LOW LIGHT
The compromise in a shot like this, taken with an 18mm EFL lens, is between depth of field, subject movement, and noise, affecting aperture, shutter speed, and ISO. Of the several possibilities, I chose f/5.6 with careful focus, 1/10 sec and ISO 400. In other words, I favored low noise, and shot many frames to guarantee the final image.
USING MOTION BLUR DELIBERATELY
Turning the problem around creates new opportunities. In this situation, of priests at prayer in a Shinto shrine, the camera was mounted on a tripod and the shot was framed to include a significant part of the setting (which remains sharp), while the white-clad figures flow insubstantially in a one-second exposure.
Because low light photography is constrained somewhat by performance limits, shooting always involves compromise, and adjusting the camera settings to sacrifice one technical function in favor of another. These are the trade-offs, and it’s easy to see them as a permanent set of difficulties, but practically it usually works better to treat them more positively, as ways of getting you closer to images that are only recently possible.
In a wider sense, all good photography pushes limits of one kind or another. Whether these are creative limits, like composition, or technical limits (as in low light) is not a great distinction, because each affects the other. Pushing the technical limits of shutter speed and aperture, and having the ability to handhold steadily and choose the right settings for the situation ultimately contributes to the creative success of an image. At this edge of technical acceptability there are bound to be failures and disappointments—perhaps many for each shot—that make this edge-pushing worthwhile. If you feel this, and are happy living with the uncertainty, low light photography will give you a permanent level of excitement.
The following factors affect the shutter speed:
How fast the subject is moving
The angle of movement to the camera: head-on, side-on, or diagonal
Distance from the camera
Focal length of the lens
In addition, the following set limits to the speed that is actually possible:
The need for good depth of field (from a smaller aperture)
The amount of acceptable noise
The amount of acceptable motion blur
The amount and direction of light
Size of reproduction or viewing of the final image
By definition, the quantity of light is always much less than the reference standard for photography and human vision alike—the sun. Midday sunlight, clear and unfiltered by clouds or haze, produces around 100,000 to 130,000 lux (see box).
The lighting in a typical office is probably between 200 and 400 lux, so the difference between the two in practical photographic terms is about EV 7 or 8. Put even more practically, at ISO 100 a typical setting for bright midday would be 1/125 sec at f/16, while in an office you would be shooting at, say, 1/15 to 1/30 sec at f/2.8 at the same level of sensitivity.
One consequence of this is that in low light there is no typical, easily predictable setting. Levels vary hugely, as the table opposite shows. Not only this, but the illumination from all local light sources, as we’ll see on the following pages, tends to be uneven. The illumination fades rapidly with distance from each lamp. The most common artificial light sources, which we’ll examine in detail in Chapter 2, are vapor-discharge lamps of several varieties, fluorescent lamps, and incandescent lamps. Many interiors and night-time exteriors have a mix of these light sources, which under some circumstances can create interesting and attractive visual effects, but which can also be a headache for tonal and color balance.
Light sources vary not only in intensity but in the color of the light they emit. The human eye manages to accommodate these differences so well that for the most part we are unaware of any major difference, other than perhaps a “warm” appearance from incandescent lighting and a slightly “cool” effect from fluorescent lamps and vapor-discharge lamps. However, the camera’s sensor captures the true emission, unfiltered by our complex eye-and-brain processing.
FLAME
Flame is the basic, original artificial light source, and still features occasionally in low light photographic situations.
TUNGSTEN
The range of tungsten lamps for domestic use is typically between 40 and 100 watts, and in color temperature between 2750 K (40 watts) and 2850 K (100 to 150 watts). However, fittings such as lampshades reduce and diffuse this output, and introduce color shifts.
CFLS
CFLs (Compact Fluorescent Lamps) are becoming increasingly common in domestic interiors for their efficiency and low heat output. However, they cause difficulties in color rendering in photography, because they lack a full spectrum.
Low light levels
The fundamental measurement unit of scene brightness is the lux. This is one lumen per square meter, and measures the amount of light falling on a scene. It has nothing to do with the light reflected from surfaces, takes no account of what is in the scene, and is akin to an incident light reading (as taken with a handheld meter). The unit more familiar to photographers is EV, or Exposure Value, and the values given here relate to surfaces in the scene, and assume average reflectivity. EV varies according to the sensitivity to which the camera is set, and the single figures given here assume ISO 100. Doubling the ISO would increase the EV by 1. EV can also be expressed more practically as a combination of shutter speed and aperture, and the examples given here are only one of many combinations. Thus, EV 7 at ISO 100 could be 1/15 sec at f/2.8, or 1/30 sec at f/2, or 1 sec at f/11, for example.
Lighting situation
Lux (light falling on scene)
EV (Exposure Value) at ISO 100
Typical speed & aperture at ISO 100
Television studio
1,280
EV 9
1/60 sec f/2.8
Indoor office
320–640
EV 7–8
1/15–1/30 sec f/2.8
Dark, overcast
80–160
EV 5–6
1/4–1/8 sec f/2.8
Twilight
10
EV 2
2 sec f/2.8
Deep twilight
1
EV -1.3
20 sec f/2.8
Average full moon
0.3
EV -3
1 min f/2.8
Average quarter moon
0.03
EV -6.5
12 min f/2.8
The illumination from any light source falls off with distance, which is fairly obvious if you think of a flashlight aimed out into the dark, or the way a room looks when only one table lamp is switched on.
In daylight photography this is not a consideration, because of the great distance of the sun. Any difference between the sunlight reaching a mountain top and that at sea level is insignificant compared with the 93 million miles between the sun and earth. Photography by any kind of artificial lighting, which covers the majority of most peoples’ low light shooting, has to deal with a very different distribution of illumination.
The light fall-off from a point source of light, such as a bare bulb, follows a simple law. It fades in inverse proportion to the square of the distance, hence the name the Inverse Square Law. So, for example, the illumination two meters away from a lamp is four times less than at one meter (1 x 1 m = 1, 2 x 2 m = 4). Put more pragmatically, light falls off rapidly, and the result in city streets at night and interiors with mainly spotlighting, such as restaurants and clubs, is a lighting pattern of pools of light surrounded by shadows. More lights add to the complication of the lighting design, and broad area lights such as diffuse, concealed ceiling lighting, lessen the fall-off. The net result is more often than not a high degree of local contrast.
This is one of the most basic technical issues in low light photography, and many of the techniques in this book are aimed at dealing with it. High contrast means not only that the dynamic range of a scene may be beyond the ability of the sensor to capture it in one exposure, but that there are likely to be significant shadow areas where detail is lost and where there is visible noise. The situation is made even more difficult when light sources themselves appear in shot—street lamps, downlighters set into ceilings, neon display advertising, and similar. The traditional ways of dealing with bright pools of light and dark shadows focus on composition and viewpoint, such as moving the camera to hide a bright spotlight. Composition can also make use of the chiaroscuro effect of light and shade, and to an extent some area or areas of dense shadow and lost, burnt-out highlight can be acceptable in the image.
HIGH CONTRAST
Sunlight streaming through an array of small triangular openings in a Sudanese tomb create an unusual and attractive effect. When contrast is very high, yet the darker areas dominate, the best strategy is to let the highlights go.
Nevertheless, digital techniques, both in capture and post-production, offer another way out of this situation, by effectively increasing the dynamic range. HDR Imaging, for example, in which a sequence of exposures is captured, then compressed into a single image file, covers a very wide dynamic range. Shooting Raw also potentially offers a higher dynamic range, although it depends on the exposure at capture as a completely clipped highlight is not recoverable. There is also the Shadows/Highlights control in a number of editing packages. Either way, post-production offers an increasing array of smart techniques for doing the two things that were virtually impossible on film—restoring highlights and opening up shadows.
To the eye, a living space like this, partly lit by subdued daylight, and with white walls helping to even out the illumination, appears reasonably balanced. A photograph, however, reveals the concentrated high illumination from spotlights and lamps, which tend to burn out in the image. New techniques of blending a dark and a light exposure, however (see here), achieve a more balanced effect that represents more accurately our visual impression.
Standard
Improved
In order of priority for human perception, the midtones of an image are naturally the most important as they usually contain the most detail, but they are followed closely by highlights, with shadows in last place.
Apart from light sources themselves, such as lamps, which are expected to be super white, we tend to sense something wrong in an image where the highlights are well overexposed. Dense shadows without features are more easily tolerated. This, more than anything, is the basic rationale for exposing to hold the highlights.
Nevertheless, there is a range of post-production techniques for recovering detail in highlight areas. All are premised on there being some detail captured. As we’ve seen, a completely blown highlight (255, 255, 255 in RGB) is simply empty space in the image. Once a photosite has filled up completely, there is no recovery possible. If one channel has blown, and sometimes if two channels have blown, there is some recovery possible in Raw, but this is the limit.
The Raw converter is indeed the place to begin highlight recovery if you are shooting this format. The more advanced image editors have algorithms for attempting to reconstruct image detail from the remaining channel or channels. Lightroom and Aperture, for example, have a specific Recovery slider for this. Dragging it to 100 percent will also affect midtone areas, but these can be pulled back with either or both the Fill Light and Brightness controls.
With TIFFs and JPEGs, the most basic technique is an adjustment in Curves. For maximum control first sample the areas of highlight that you want to alter, in order to define exactly which section of the curve to drag. Anchoring other parts of the tonal range by selecting points on the curve below and even above this section will help to constrain the effect. However, curve adjustments tend to carry some unwanted side effects, like lowering contrast. Some recent methods are starting to make Curves look primitive.
These methods of altering tonal range take account of the spatial location of each pixel, and are a form of local operator. Probably the best-known is the Shadows/Highlights control found in a number of programs. However, extreme settings can produce unrealistic results.
In this shot of firecrackers being thrown into a furnace in a Chinese temple, the flames are very overexposed, but recovery of sorts is possible.
Using Lightroom delivers an improvement, achieved by increasing Recovery to the maximum, decreasing Exposure and increasing Brightness.
In Shadows/Highlights, Shadows are turned off, as we only want to recover the highlights. The first step is to find the tonal width that covers all the flames in the furnace opening. To do this, the Highlights amount is set to maximum, and the Tonal Width slider moved around to find the best effect. The Radius slider is also moved to find the most pleasing effect. The Highlights Amount is then adjusted for best effect.
As with highlight recovery, if your exposure has been made for the midtones, as is usual, there will often be a need to lighten the shadows in a high-contrast image.
As just explained, dense shadows are perceptually more acceptable than washed-out highlights. At the same time, there is usually more that can be recovered, as there are usually some photons of light that have entered even the darkest photosite on the sensor. Practically, this means that it is usually easier in most images to pull more detail out of shadows than out of highlights. There is an important limitation, however, which is that the most noise is in the shadow areas, and any enhancement of shadow detail will exaggerate this noise. In the darkest areas, the signal-to-noise ratio may be so low as to mark the limits of useful image. So, any shadow recovery technique usually benefits from noise reduction as well.
The techniques are similar (though in reverse procedure) to those for highlight recovery, beginning with the Raw converter. In Lightroom, the relevant slider is Fill Light. Understandably, it has a slight flattening effect on the dark-to-mid tones, and may benefit from some contrast enhancement. The upper set of sliders in Photoshop’s Shadows/Highlights control work as the Highlights do (see here).
Finally, it’s worth mentioning in this context another local contrast enhancement technique that is really an unintended by-product of sharpening. This extremely useful technique is applied through the standard sharpening method known as Unsharp Mask (USM). While seemingly different in purpose, sharpening and local contrast enhancement are very closely related. Both depend on creating a soft mask, across which the contrast is increased. In USM, the contrast increase is on a very small scale—often a radius no more than one pixel—but quite strong, so that edges become more pronounced. In local contrast enhancement, the settings are reversed, with a very large radius, often as much as 100 pixels, but a small amount. The effect is larger-scale transitions, but without altering the global contrast of the image, and is typically subtle. In both, some pixels cross over each other in the histogram, making this completely different from a curves-based adjustment. The latter, one form or another of an S-curve, is a truly global adjustment.
Other image-editing software offers alternative techniques for opening up shadow areas, generally through either lightening a selected area or by means of tone mapping and accentuating local contrast. The two programs featured here are both aimed at photographers in the sense that they deliberately avoid the Photoshop way of working. In different ways, they attempt to be intuitive and automated, so as to speed up processing decisions. In the course of doing this, many of the arcane parameter sliders and tools common in Photoshop are concealed, or at least pushed into the background. This applies to shadow control as much as other procedures. DxO Optics Pro has very little showing in its controls that appears to be specifically aimed at opening up shadows, but the algorithms are there nevertheless. They are included in DxO Lighting tools, and the principle they use is segmentation, in which the image is analyzed for different ranges of luminance values, and these are then each treated slightly differently.
Lightroom default
Lightroom with Fill Light at 25
Before: LightZone offers another option, Relight, which performs a similar procedure to DxO.
With Relight
Original
DxO Open Shadows
DxO Preserve Shadows
DxO uses the reverse option of Preserve Shadows, which holds them dark. Without the option ticked, they receive tone mapping to open them up.
Shadow areas are generally given an increase in lightness and the tone curve slope. The choices Slight, Medium, and Strong then apply these, as well as other automatic adjustments, to different strengths accordingly. In addition, unchecking the Preserve Shadows checkbox opens them up even more, at the slight risk of a loss of naturalness.
LightZone offers more than one method of opening up shadows specifically. One is the Relight tool, which uses tone mapping techniques similar to those seen in high-dynamic-range software to lighten shadow areas while maintaining the local contrast in the image (and the reverse for highlights).
Another is LightZone’s basic Zone System approach to tonal adjustment, whereby a zone can be selected and dragged up or down in the scale to lighten or darken just those values. Yet another option is the distinct High-dynamic-range/Dark-scene “style,” one of a number of presets that opens up a tool giving options for adjusting shadows.
Photoshop’s post-processing option applies tone mapping selectively to darker and lighter areas.
This very gentle effect is achieved by first creating an inverted Luminosity mask, and then applying USM at low intensity and high radius.
Original
Clicking this symbol creates a luminosity mask of the lightest 50 percent.
Invert the selection to select the darkest 50 percent of the image.
USM applied to subtly add contrast to the shadows only.
Over the last six pages we viewed a number of powerful post-production tools for recovering lost detail and enhancing limited detail at either end of the tonal scale.
Because of the nature of low light photography and its typical situations, shadows play a very large part in the image. The efficiency of image-editing software in pulling detail out of shadows makes it tempting to use aggressively, but this carries some dangers. One obvious problem is the enhancement of noise. Even when noise is fairly evenly distributed across an image it is not particularly attractive. When it is concentrated very noticeably and patchily in the shadow areas, it is even worse.
But beyond this, there is the more subjective issue of what looks right in an image. This is impossible to quantify, because it varies from viewer to viewer, and is affected not only by what we consider to be the normal, realistic appearance of a scene, but also by the way we have become accustomed to seeing photographs. In a sense, shadow recovery techniques are a small but important part of the much bigger issue of tone mapping HDR images. The effect that shadow recovery has is the appearance of more visual information, but not necessarily a more truthful, convincing, or even aesthetically pleasing image. The sequence of image adjustments shown here makes this obvious. As the recovery procedure is applied more intensely, the shadow areas throw up more and more detail, but at a certain point they begin to look false. Exactly which step looks false to you is a matter of personal judgment, but it has become one of the most common and glaring “faults” seen in today’s digital image processing, especially with low light pictures.
JUDGING THE EXTENT OF TONE MAPPING
Using Aperture’s Highlights & Shadows for shadow recovery only, this series of different amounts shows that while increasing the number of opened-up shadows obviously reveals more and more detail, at a certain point the image will begin to look artificial and lacking in traditional photographic qualities.
Color temperature is arguably more important in low light photography than in any other, because the range of light sources you are likely to encounter varies widely on this scale.
First though, with apologies to all those people who know this already, a brief primer on the color temperature scale, and the reasons why it has a place in photography. It all starts with the sun, which is the reference standard in several ways for human vision. Its light, at least when it is high in the sky, is what we call white. As this light is created by burning—incandescent, in other words—there is an exact correlation between temperature and “whiteness.” The unit of temperature used is the Kelvin, which is basically similar to Celsius/Centigrade but starts at absolute zero (-273ºC). Something burning at a lower temperature than the sun, such as a candle, produces a redder light. Something burning at a higher temperature (few things on Earth practically, but a hotter star for example) will produce a light that is more blue. When most artificial lighting was incandescent—tungsten lamps—it made sense to use temperature as a color scale, from red through orange, amber, straw and white to increasing blue.
PLANCKIAN LOCUS
The color temperature scale plotted on a CIE 1931 x, y chromaticity space.
COMPROMISED
BALANCED FOR OUTSIDE
BALANCED FOR INSIDE
SUNLIGHT
Sunset colors are created by refraction and scattering, and here cover the apparent range from about 3500 K to less than 1500 K. The actual color temperature of the sun’s disc is under 6000 K.
It makes less sense now that artificial lighting is increasingly dominated by fluorescent, vapor discharge and even LED lamps, but because photography has grown up using color temperature and Kelvin, these new light sources are made to fit into it the existing system. More accurately, they lie on a correlated color temperature scale.
The diagram shows what fits where. To be completely accurate, only incandescence really belongs on this scale, meaning the sun, an incandescent lamp such as tungsten or quartz-halogen, and a flame from a candle or a fireplace. Even the reddishness of the sun at sunset and sunrise, and the blue of the sky, are not strictly speaking on the scale, although they appear to fit very well. They are caused by the scattering of short wavelengths of light. Fluorescent and vapor discharge lamps only just fit at best, as they emit light by very different means, and they certainly need additional hue correction. To say that a typical fluorescent lamp has a color temperature of 4200 K, as does my camera instruction manual, is quite misleading and only partly useful.
Beware of confusing the natural, visual way of describing this kind of color shift with the real color temperature scale. To the eye, the lower color temperatures associated with flames and sunsets are “warm,” while the higher color temperatures that make evening skies and open shade look bluish are what most people call “cool.”
White balance in digital photography, as its name suggests, is the procedure for adjusting/correcting the overall color cast of lighting on a scene.
The principle is that by identifying areas that should be white in the image, then making adjustments so that they appear true white without a color cast, the color of the light falling on all other areas of the picture will also be neutralized. The obvious place to do this is in the camera as soon as the image has been captured, and this is practical because the color component of the image needs to be calculated in any case. As we saw before, the Raw capture is of tonal values only—effectively a monochrome image—and the color is sensed by filtering these values, pixel by pixel, through a mosaic of red, green, and blue. This mosaic, the Bayer array, has to be interpolated from neighboring pixels. It is relatively straightforward to bias this interpolation toward different color shifts. Naturally, this is more important in low light photography because of the wide range of possible light sources.
This could actually be done in any number of ways, but because photography has grown up with color temperature, this is the scale used in all cameras. However, as we just saw, many light sources lack a continuous spectrum, and technically they do not really fit onto the scale of color temperature as measured in degrees Kelvin. But, because color temperature is such a familiar and convenient concept, they are made to fit approximately. This explains, as we’ll see here, why the camera menu choices for fluorescent lighting frequently appear not quite right. Most good digital cameras also offer some adjustment to each white balance choice. This can be to raise or lower the color temperature a little, or to shift the colors around the color circle slightly. Check the camera manual to familiarize yourself with the method used, as this varies by manufacturer.
Shooting Raw, which is recommended strongly for low light photography, makes the white balance set in the camera irrelevant, as you can choose the settings with absolutely no loss of quality on the computer during processing. Raw converters, of which there are many, give not only a scale of color temperature, but also of hue. This is basically a second color axis more or less perpendicular to the color temperature scale, and the two together give a huge amount of possibility when it comes to color adjustment.
Technically speaking, choosing the appropriate white balance setting for a particular image means neutralizing the color cast—making corrections so that it appears as if pure white light is falling across the entire scene. However, slavishly following this is not guaranteed to produce the most pleasing results. Not only are color casts not bad, they often contribute very much to the final color image.
A dropper allows the white balance to be set by choosing any tone that should be neutral (not necessarily white). There may be subtle, but noticeable, differences even between apparently similar neutral areas.
The white balance settings in the camera menu and, for Raw images, in a Raw converter, adjust color temperature and hue. These four versions of white crockery set on a white Plexiglas display unit lit from beneath by unknown fluorescent lamps, were originally camera presets. They are, however, accessible and amendable in a Raw converter if the image was shot, as here, in Raw format.
In-camera Auto
Lightroom Daylight preset
In-camera Fluorescent
Lightroom White Balance Dropper