Capturing detail in low light is one of the biggest challenges for any camera, but it is possible to overcome this difficulty and achieve stunning results.
Recent advances in digital imaging have seen manufacturers improve how cameras perform in poor lighting conditions. However, even when using the best cameras currently available, it is still vital to correctly expose the image to minimise noise and ensure maximum image detail.
When shooting in low light, getting the correct exposure for the subject is key to crisp, saturated and dynamic images. The ‘best’ exposure settings are highly camera specific.
Unfortunately, many of us switch to autopilot when shooting in low light – simply increasing the ISO rating without giving much thought to noise or shadow detail. However, some basic techniques can help to improve images taken in low-light conditions, particularly at high ISO settings.
In this article, we will explore how to avoid the impact of noise in low light using a series of tests. Different cameras have been studied, and by understanding and implementing the results of the simple test on the opposite page, you can discover the ideal exposure settings to produce the best results in low light.
But before we do this, here are a few things that will help you to understand your camera and its sensor.
The trouble with understanding ISO ratings in digital photography is that the term ‘ISO sensitivity’ originally comes from film photography, and there are differences between film and digital capture methods. The chemical make-up of a film base affects its sensitivity to light, with a higher ISO rating being more sensitive. A digital imaging sensor does not have an ISO rating. In fact, the sensor’s ‘sensitivity’ to light is constant and is in no way affected by the ISO setting. The quoted ISO sensitivities found on cameras are, in fact, ‘equivalent’ settings. Rather than the sensor actually becoming more sensitive to light, all that happens is the development process (raw sensor output to final digital image) adjusts the brightness and contrast to produce an image with a correct tonal range.
A simple analogy is that a digital sensor is like shooting on ISO 200 film and then changing the camera settings as if using ISO 3200 film. The sensitivity of the film doesn’t change – it is still ISO 200, despite what the camera’s exposure settings say. To be able to use the now underexposed film, it must be pushprocessed and printed appropriately to get usable images.
A silicon digital camera sensor yields an output directly proportional to the level of light that it receives. So, to increase a digital sensor’s exposure to light, one must either increase the strength of the light source, the length of the shutter speed or the size of the aperture. How much light is collected and used by the sensor depends on its quantum efficiency (see right).
As the sensor doesn’t become more sensitive as the ISO setting increases, it is advisable to give it a brighter exposure than the nominal one that the camera recommends.
The lower the sensor’s exposure to light, the greater the level of noise will be. Therefore, by giving the sensor more light, the level of noise can be reduced and the tonal range can be improved.
One way to give the sensor more light is to expose the scene so that the brightest point is exposed as much as it can be without clipping. In many cases this will cause the overall image to look overexposed. However, by increasing the exposure from the nominal setting (which was suggested by the camera), the sensor receives more light and therefore produces less noise and more detail in shadow areas. The image can then be adjusted in editing software to reduce its brightness, with the added advantage that any noise that does still exist in the image will also be darkened, making it less visible.
So, regardless of the ISO setting, overexposing an image up to the point of highlight clipping, and then reducing the brightness in software, should help reduce noise and improve detail in shadow areas. In summary, the ‘best’ exposure is the brightest before highlight detail is unrecoverable.