[Tig] The most out-of-gamut colour
richard at filmlight.ltd.uk
Wed Feb 1 17:19:32 GMT 2017
Here's an answer to Sam Sheppard <sam at silverscreenpictures.co.uk>.
The P3 triangle of colours contains most of the possible bright reflection colours. You could make something that gave a very saturated reflection colour. A matte version of this colour would be very dim because a broadband white would only have a small fraction of its energy in a particular band. A specular reflection colour, such as a peacock tail feather, may reflect brightly in a particular direction, but the colour you will see will depend on the exact viewing angle. We know that peacock tail feathers can look blue, or green, or dark when we turn them. We know iridescent colours do not behave like normal pigments, and we do not expect them to have a consistent hue or brightness.
We might expect a narrow wavelength light source such as a LED or a sodium street lamp to have an exact colour. Again, they do not quite do this. Our brains are finely tuned to interpret colour under broadband illuminants: so much so that many people consider 'colour' to be an attribute of an object, and not a property of the object and the illuminant. However, there are self-luminous objects such as car lights at night, traffic lights, laser pointer green - we should try and get those colours right, right?
Again, there is another level of problems. We see colours via our LMS cone detectors, and ganglion. These network of non-linear processes give us a sense of hue and brightness, but the hue signal depends on the intensity. This is not normally a big deal, because all the cunning stuff in our heads that gives us colour constancy manages to correct for much of this.
I have cut a lot of detail here. But the general upshot is there is not nearly the same range of useful different shades outside the typical P3 triangle as the areas of any u'v' plot would have you believe. The Rec-2020 gamut gets in some more colours, but are those colours useful? If your white is made of more saturated colours, and everyone sees precise wavelengths slightly differently, a bigger triangular gamut may make us see different whites. A Rec-2020 green may be good for rendering a laser pointer of that exact colour, but it sits at the very pointy end of a triangle, and it is not good at mixing the shorter wavelength colours. I think the P3 gamut contains not only the Pointer gamut of reflection colours, but most of the non-reflective colours that we can reliably and consistently see.
So, what do we do with real images and a real display?
Baselight 5.0 has a strip called Compress Gamut. This will compress all the possible colours outside the RGB triangle to lie within the RGB triangle To do this, it compresses the colours towards the edge of the triangle a bit to make room, but the colours that lie below (say) 80% saturation are untouched. This means most of the colours in the standard video gamut re untouched - and the only ones that are touched are probably near the edge of the video gamut and were risking clipping anyway. This filter will not give you the missing 500nm blue green colour, but it will find you a less saturated hue match, so you do not slam into the RGB limits one at a time, and get mach bands.
There are other tools, but this is probably our best one for stomping a big gamut into a small one. I don't know what other systems do.
FilmLight Ltd, Artists House, 14-15 Manette Street, London W1D 4AP
Tel: +44 (0)20 7292 0400 Fax: +44 (0)20 7292 0401
More information about the Tig