numerical methods - sRGB's linear segment to avoid infinite slope - why? -
in article srgb (https://en.wikipedia.org/wiki/srgb) stated, gamma transformation has linear portion near zero, "avoid having infinite slope @ k = 0, can cause numerical problems". i'd know what's problem that.
there 2 answers, usual gamma. modern variant is:
the problem infinite slope need "infinite" resolution (many bits of storage) in order arrive @ linear representation invertible gamma-encoded without loss. in other words, allows small lookup table produce invertible linear encoding (8bit -> 10 bit -> 8 bit).
the numerical problem understood on first step (8 bit -> 10 bit). infinite slope near zero, need bigger encoding range stay faithful , reversible, i.e. you'd need more 16 bit (assuming integer coding, halfs not have problem).
the linear equivalent of #010101 or 1/255th square (gamma = 2.0) coding 1/(255*255)th. need 16 bits represent faithfully, , using 2.2 not 2.0 exponent make worse. these quite small numbers corollary of coding function, , in practice don't need resolution in lightness range is, roughly, black. linear segment helps coding not wasting resolution detail around black (or near zero).
the older answer, taken from
http://www.poynton.com/notes/colour_and_gamma/gammafaq.html#gamma_correction
is in equipment linear segment less sensitive noise. true of analog signal path.
Comments
Post a Comment