Noise is the digital equivalent of grain in film. It’s random electrical signals captured by the photosites on the camera sensor, and usually this background noise level is so low compared to the brightness of the captured picture itself that you just don’t notice it.
But if you start increasing the camera’s ISO setting, the noise becomes more prominent. It’s like turning up the volume on a quiet analog audio tape recording – the sound gets louder, but so does the background hiss which you never noticed before. In other words, the signal to noise ratio starts to get worse. Increasing the ISO is often the only way to get pictures in very low light situations, such as this pre-dawn shot of a fisherman.
The larger the photosites on the sensor, the better the signal to noise ratio. This means that physically larger sensors generally produce less noise than a smaller one. This is another reason why sensor size is important.
The sensors in smartphones, action cams and point and shoot cameras are so small that there may be visible noise even at their lowest ISO (sensitivity) settings.
On a camera with a small sensor you might hesitate to use an ISO setting of 400 or above, while on a full frame camera you might be able to get very good image quality at ISO settings of 6,400 or higher.
This means camera and software makers work quite hard at noise reduction. Some cameras have noise reduction circuitry built into the sensor, most apply some kind of noise reduction when processing photos, and almost all photo editing software has noise reduction tools.
Noise reduction is not a perfect science, however. Very often, the noise reduction process does more harm than good, removing the hard, pixellated granular noise in the image but replacing it with a smoothed-over ‘watercolor effect’ that leaves fine textures smoothed over and lacking in detail. This is a persistent problem with small-sensor cameras.
Luminance vs color noise
Digital camera noise comes in two types: luminance (contrast) noise and color noise.
Color noise is random variations in color between pixels and it’s quite easy to remove without harming the appearance of the image. You rarely see it these days because it’s processed out in the camera or, if you shoot RAW files, by the default settings in your RAW processing software.
Luminance noise is much more problematic. This is random variations in brightness between pixels, and if you smooth these out you start taking away the find detail in the picture. Despite what the makers say, even the best noise reduction processes today still tend to remove or smooth over real image detail at the same time as removing luminance noise.
The advantages of shooting RAW files
If you shoot JPEGs with your camera, you are stuck with the camera’s own internal noise reduction processing. Sometimes this can offer just about the best compromise between noise reduction and detail rendition, and often you can change the noise reduction settings.
Usually, though, you can get better results by shooting RAW files and then choosing the best RAW processing software and noise reduction settings. Some RAW converters are better than others. Lightroom and Adobe Camera Raw are not particularly good at noise control and detail retention, while DxO PhotoLab Elite’s PRIME noise reduction is a class above the rest – though processing is slower.