Exposure and Workflow
By Bob Newman, first published December 2012
In the last two articles the topic of discussion has been exposure, which is defined as the illumination of the sensor times the time for which it is illuminated. Exposure determines the density of light incident on the sensor over the whole exposure time, and thus the number of photons, which make up the image. From that number of photons is derived the photon shot noise, the major noise in an image, so that exposure directly controls image noise – the greater the exposure the less noisy the image.
Thus, if the photographer’s aim is to produce images with as little noise as possible, which would be a common view of what constitutes the highest ‘image quality’ then the aim will be to manage exposure so as to minimise that noise, which in turn simply means ensuring that the exposure used is as large (that is, as much light, in lux-seconds) as possible. There are of course limitations to this approach. Always there are factors which stop a photographer choosing an arbitrarily large exposure. One is the exposure which the camera can record. Another is control of motion blur, which limits the shutter speed. Aperture is limited by the size of the lens, or by the depth of field required. So, in the end the art of managing exposure consists of juggling the desire for the largest possible with these other needs, which I will collectively call the ‘pictorial constraints’, that is they are constraints born out of the desire to take a particular picture, rather than any other.
The other set of constraints on choice of exposure are ones which I would call ‘workflow constraints’, where for the purposes of this discussion there is a very wide definition of ‘workflow’, starting with the way that a photographer chooses to use a camera to the output medium, covering between those the development and post processing employed.
In camera processing and film emulation
So long as a photographer sticks to the in-camera processing, a modern digital camera behaves very like an old film based camera, being particularly close to a camera using transparency film. This is not because of the characteristics of the underlying technology but because they were designed that way so as to be easier for photographers raised on film to use. Figure one shows a series of characteristic curves, which plot output brightness against exposure.
Figure 1: Film and JPEG react to light in a similar way. The same exposure change produces much less effect in the ‘head’ and ‘toe’ of the S-shaped curve than it does in the centre. By contrast, the same exposure change causes the same effect to the data in a raw file, except right at the top, where the sensor begins to saturate.
The similarity of camera processed JPEG files to film can be seen from the similarity of the film curve and JPEG curve. The dissimilarity of the JPEG curve to the underlying characteristic of the sensor can also be seen. Whereas the JPEG and film curves are ‘S’ shaped, and compress the highlights and shadows with respect to the mid tones, the raw curve is linear, with even response to all brightnesses. The shape of the S curve depends on development. If using out of camera JPEGs, it is the camera’s firmware which determines the shape of the curve and which part of the linear sensor characteristic goes to make the final brightness curve. This is the way that a range of ‘ISO’ ratings are produced from a single sensor output characteristic – the camera selects to use different sections of the sensor curve to map into the tones of the final image, as illustrated in Figure 2.
Figure 2: Different ISO curves are created by mapping different parts of the raw characteristic to the desired ISO S-curve.
The consequence of this is that if using a workflow based on ex-camera JPEGs, we need to take into account the processing characteristics that were previously selected, by dint of choosing an ISO. On the other hand, when shooting raw, the processing will be post-selected, so the processing can be optimized for the exposure used, rather than the other way round. In this case, the photographer has rather more latitude with respect to choosing exposure parameters for pictorial effect, such as depth of field and motion blur.
JPEG workflow
One of the great attractions of digital photography, at least for photographers brought up on film and used to processing their work in a commercial laboratory, is that a digital camera using JPEG is much like a film camera loaded with a huge roll of film, that can be erased and reused. This is because the in-camera processing has been designed to react to changes in exposure just like film did. Looking again at figure 2, we can see that the change in output brightness with respect to exposure is not linear. The same change in logarithmic exposure produces a much smaller change in brightness at the head and foot of the curve than it does in the centre. It is this characteristic that allows typical display media such as computer, TV and phone screens and print paper, all of which are capable of displaying tone with a brightness range of eight stops or so, to display a wider range of tones, which might occur in a real life scene. In the shadows and highlights, the tones become compressed, due to the shape of the curve. Subjectively, this is all right, since we will tend to concentrate attention on the md tones, we don’t expect to see a great deal of detail in shadows or highlights – but we do expect to see some. Thus, the S-shaped curve provides a more convincing rendition of reality than would a simple eight stop linear curve. Fortuitously, this is the characteristic the produces naturally – it is unsurprising therefore that the engineers designing digital cameras decide to use it.
When using JPEG or film it is necessary to choose the exposure so that the local exposure generated by different shades in the scene match the correct parts of the S-shaped exposure curve. The cost of getting this wrong is that the most prominent parts of the scene end up being recorded in the squeezed head and toe parts of the curve. Although the brightness can be returned to nominal using post processing tools, much information on the tonality in the mid-tones will have been lost due to the compression, and the result is often an unsatisfactory photo. Thus, when using a JPEG workflow, there is a big incentive to ‘get it right in the camera’, that is to take care adjusting the exposure to ensure that the outcome is positioned correctly on the S-curve for the ISO selected.
Raw workflow
Things are slightly different using a raw workflow. As can be seen from the graph in figure 2, the raw output of the sensor is linearly related to the exposure level. When you view a processed image from a raw file, the film-like S-shaped curve is imposed during processing, and processing can be adjusted to suit the latent captured image in the file. Thus, in principle, it matters little on which part of the curve the exposure lies, when correctly processed the full tonality will be available. The limit to this is the acceptability of noise. As seen in previous articles, the lower the exposure, the higher the noise. Thus, using a raw workflow, the aim of exposure management is to maximize exposure, rather than attempting to match the S-curve – which instead can be matched to exposure. There are a number of factors which limit the size of exposure that might be used. The first is the extent of the raw curve, if the exposure is so large that it goes over the top, the highlight detail will be lost. Next is the limit on shutter speed, which is usually a matter of minimizing camera shake, and lastly limitations on f-number, either the maximum aperture of the lens, or the required depth of field or in some cases, the desire to use the lens at its optimal sharpness. Conventionally exposure is set by the ISO selected, so this advice amounts to selecting as slow as possible an ISO. Non-conventionally, we have seen in previous articles that exposure and ISO, in manual mode (or using exposure compensation) are not as tightly linked as is often suggested, that is there is leeway to use a larger exposure than the nominal one for the ISO. The result of doing so is lower noise, the risk is that some highlights might be lost, by straying over the top of the characteristic curve. One way to check this is to look at the histogram on the LCD display (either after a test shot or in live view, if a histogram is available) , and increase the exposure ensure that there the clear space to the right is minimized, but existent – a technique known as ‘expose to the right’.
Bob Newman 2024
