The role of gamma correction in colour image processing

In this paper we analyse the influence of gamma correction of digital colour cameras on computer vision algorithms. These nonlinear signals R', G', and B' yield a nonlinear shift in colour space models and can cause algorithms to fail. Experiments of colour segmentation and tracking confirm the claims. We suggest to use linear light signals R, G, and B for calculating most colour space models instead of the nonlinear signals R', G', and B'. To obtain linear light signals either the camera has to be configured linear using the gamma switch (if available) or by removing the nonlinearity analytically using the inverse operation (gamma re-correction).