This usually refers to a 10-bit sampling system that maps analog values logarithmically rather than linearly. It is widely used when scanning film images which are themselves a logarithmic representation of the film’s exposure. This form of sampling is available directly from some digital cinematography cameras.
The 10-bit data can describe 210 or 1024 discrete numbers, or levels: 0 – 1023 for each of the red, blue and green (RGB) planes of an image. However, as all electronic light sensors have a linear response and so produce an output directly proportional to the light they see, when scanning film they represent the transmittance of the film. Usually it is negative film that is scanned and this means a large portion of the numbers generated describe the scene’s black and dark areas (representing bright areas of the original scene), and too few are left for the light areas (dark scene) where ‘banding’ could be a problem – especially after digital processing such as grading and color correction. Transforming the numbers into log (by use of a LUT) gives a better distribution of the digital detail between dark and light areas and so offers good rendition over the whole brightness range without having to use more bits. A minimum of 13-bit linear sampling converted to 10-bit log sampling means sufficient detail in the pictures is stored to allow headroom for downstream grading that is common in film production.
10-bit log is the basis for sampling in the Cineon and SMPTE DPX formats that are still widely used in the post production and DI industries.
See also: 10-bit lin, LUT