Colours at high ISO part 2, Signal to Noise

Image of Mbulelo Ndabeni dancing in hoodie, his face expresses something between concern and terror. He is holding a slatted bench at an angle and appears to be peering out at the camera as if ready to hide behind it or run away at any time. Image has little or no detectable noise (speckling) despite being taken at ISO 3200.

Mbulelo Ndabeni performing iBali – Echoes of the Future. ISO 3200 and 1/80 second.

Modern digital cameras handle ISO incredibly well, especially when considered against the limited ISO range that used to be available to analogue (film) camera photographers and digital camera users as recently as 10 years ago. Sensors perform best at low ISOs, and as the ISO increases the image quality decreases. In the same way that film gets more grainy at higher ISO levels, using the camera’s digital sensor at higher ISOs increases digital noise.

There are two types of noise or graininess; chroma noise and luminance noise. The latter (luminance noise) retains much of the original colour while the former (chroma noise) looks like coloured speckles because the colour of the pixels has been affected. The consequence of more noise is less sharpness, less faithful or less saturated colour, a rougher-looking image and consequent decrease in image quality.

As I wrote in last week’s post, colour in digital cameras is calculated through application of a micro lens matrix. While many cameras automatically tweak the colour saturation, this has variable results and doesn’t change the limited colour palette available at higher ISOs.

Meanwhile, as the signal from light intensity decreases, a greater electrical charge is applied to make a digital camera sensor more light sensitive. Noisy light needs more processing and while a lot is removed through the processing, sometimes that noise can be more pronounced. This looks like speckles on the image.

The processor, its megapixel count, the sensor size and micro-lens matrix all play a part in how well a camera can minimise noise, and the threshold beyond which the quality of image will significantly decrease (different in different cameras) is known as the signal-to-noise ratio.

Most post-production software reduces chroma noise quite well and luminance noise less well. The best way to work out what best suits your style is to experiment by reducing each, separately and together.


ISO basics for beginners: