What is a proper exposure?
If you are taking JPEG images, then the proper exposure is the one which makes the image look best to you. JPEG images are not really suitable for image manipulation, so what you see is what you get. On the other hand, if you are capturing RAW files and processing them using computer software, then the proper exposure is that which captures high-quality image data best-suited for processing by your software. High quality RAW files don’t necessarily look good in the preview on your LCD screen.
How does your camera sensor see light?
To understand exposure, first we need to understand how the camera sees things as compared to how our eyes see. The camera sensor basically counts photons of light. Photons striking each pixel cause an electrical charge to accumulate on the sensor which is measured by the camera. Up to a point, the more photons striking the sensor, the greater the charge. The camera software assigns a number proportional to the charge which is part of the RAW file and indicates brightness at that pixel. However, there is a limit to how much charge can accumulate. Beyond that point, additional light striking the sensor can’t further increase the charge, so no more information accumulates. When an image is overexposed, any tonalities exceeding this maximum amount will be assigned the same maximum number, no matter what the tonality. This is commonly referred to as “clipping” the highlights, and is one of the worst things you can do, because there is no way to retrieve the lost differences in tonality.
How do we see light?
While the camera sees a linear relationship between the number of photons striking the sensor and brightness, our eyes see things differently. For a variety of reasons, our eyes and brain enhance the brightness of tones at the darker end of the tone curve, while compressing tones at the brighter end. Presumably this is an evolutionary adaptation that allows us to see danger lurking in the shadows, even on a bright day. A result of this is that we perceive a middle gray tone (or midtone) at about 18% maximum brightness rather than 50%.
This is important because the light meter in your camera uses this midtone or 18% maximum brightness as the target to aim for when suggesting an aperture or shutter speed. Typical objects which we photograph in nature reflect light in this range. However, as will be discussed below, the camera sensor captures more information at the brightest tonalities even though we discriminate tones better at the lower end of tonality.
What is the dynamic range of my camera?
Dynamic range refers to the ratio of brightness between the brightest thing you can see detail in and the darkest. In nature, on a very bright day, the brightest objects may be more than a thousand times as bright as the dimmest. On the other hand, some images taken under subdued light may show less than 50 times difference in brightness. Photographers frequently measure dynamic range in f-stops, where a difference in one stop means a doubling (or halving) of the amount of light. A dynamic range of 5 stops is 2 x 2 x 2 x 2 x 2, which is 2^5 or 32 times as much light from brightest to darkest. Most modern DSLR cameras can capture more than 10 stops of dynamic range, which is more than 1000 times difference in brightness. However, the image quality may be reduced at the lower stops. At low exposure levels, noise, which is the accumulation of random electrical charge on the sensor, becomes noticeable in comparison to the charge produced by the photons. This appears as graininess in the image (see How to Determine the Dynamic Range of your Camera). The resulting poor image quality in underexposed shadow areas may be more noticeable in large prints viewed up close than on smaller prints or images viewed on a small LCD screen. The best image capture is in the brightest 5-7 stops, although images capturing more stops can be processed with computer software.
The range of tonalities captured by your camera can be displayed as a histogram on the LCD screen of most digital single lens reflex cameras. The histogram is simply a graph of the number of pixels which registered at each level of tonality, with the brightest tonality on the right. Typically, only the 5 brightest stops of tonality are shown, with the midtone in the middle. In the left side figure below, a diagram of the histogram shows the relative brightness of each stop. This representation is similar to how our eyes see things, with each stop, or doubling of brightness, seeming to represent a linear increase in light. On the right are histograms for the overall luminosity (lower) and individual RGB channels (upper) for a representative image.
Expose to the Right
Thomas Knoll (designer of Adobe Photoshop), in a discussion with Michael Reichman on The Luminous Landscape website, pointed out that since the brightest stop on the histogram contains one half of the assigned levels of information, much of that available space to record information is not used when we center the exposure around the 18% midtone. Such an exposure centers the image around our perception of brightness, but not around the camera sensor’s ability to record light. For example, my camera records 14 bit RAW files, which therefore contain about 16,000 recordable levels of brightness. Since each stop corresponds to twice as much light as the next stop down, the brightest stop records 8000 levels, the next 4000, the midtone stop 2000, and so on. In the diagram below, the stops are represented linearly in the same way as the sensor records the information. Below it are shown 2 lines representing how the sensor would record an image containing 3 stops of brighter detail plus two stops of shadow detail.
In an image centered around the 18% midtone (red line), the image is recorded with fewer available levels of information, compared to an image exposed 1 stop longer (ETR, green line). In particular, the shadow areas are recorded with potentially significantly fewer levels. In addition, there will be more noise in the lowest stops. To optimize the capture of information in the shadows, it is better to record the image with the brightest pixels captured nearer to the maximum brightness (i.e. to the right of the histogram) but WITHOUT CLIPPING THE HIGHLIGHTS. On the histogram you want to be sure that the right end of the graph is not crowded against the right axis. On the left side of the figure below, the image is centered around the midtone, with very little information captured in the highest stop. Exposure could probably be safely increased by one stop. On the right, the image was exposed 2 stops longer. The data is crowded against the right axis indicating clipping and loss of highlight data. Exposure would need to be reduced by at least one stop.
Exposing to the right ensures that the shadow areas will be recorded with the best possible detail. An image recorded in this way may seem too bright on the LCD screen, but will be adjusted later using image processing software. Exposing to the right doesn’t make sense for JPEG capture, which will not be adjusted later. It is important not to get crazy about this, however. Low contrast images recorded at the center of the histogram will yield good photos, as long as there is not a lot of important shadow detail. On the other hand, clipping important highlight details needs to be avoided at all costs, because this information is lost forever. We’ll talk about how to do this in the next section.