Bias subtraction is a data reduction technique used to remove fixed-pattern noise (i.e. bias) in the image. This type of noise originates from the uniqueness of each specific sensor; even within the same model, no hardware part is ever exactly the same.
Using bias subtraction on raw images results in a bias subtracted image with a higher signal-to-noise ratio and scientific significance. This technique is used to calibrate practically every camera type and many manufacturers include the correction automatically in their images. While Nüvü Camēras gives access to raw images to suit its highly technical users, it is recommended to always use bias subtracted images.
Many factors influence fixed pattern noise in a camera, as every part of the hardware chain (sensor chip, analog to digital converter, etc.) contributes to the bias. This noise implies that even a pixel containing no charges will not necessarily output 0 volts upon readout; negative values can even be possible. To correct this phenomenon, it is common practice to add a baseline value to images; commonly known as a clamping value.
Contrarily to other types of noise, this bias has a fixed location and amplitude and can thus be removed from the image. To do so, dark frames – closed shutter images without any signal – are acquired at the maximum frame rate by setting the exposure time to zero. Minimal exposure time is used to avoid significant contributions from here). The median of ~100 such images is used to create a bias reference image, which further reduces susceptibility to extreme variations caused by readout noise or . The bias reference is removed from every raw image to create the bias subtracted image.(see more information on noise sources