"Kenneth" <kdw### [at] gmailcom> wrote:

> And getting back to the subject of gamma 2.2 non-linear blending of colors: It
> occured to me that POV-ray's focal_blur camera might be mistaken by some users
> for actual 'blurring'-- when it's really just lots of discreet 'randomized'
> camera rays, shot out in a stochastic fashion. I.e., there's no actual
> *blurring* of scene colors to cause the gamma 'darkening' effect.
>
> But that raises a question in my mind: If a focal_blur camera has it
> blur_samples set at a high value (like 500, for example), do any of those
> discrete samples actually overlap? IF so, then I would imagine that there's some
> color 'averagaing' going on behind-the-scenes, of the overlapped rays. Does that
> relate to the gamma 2.2 color-blending problem? (Hmm, probably not-- because the
> 'averaging'-- if done-- is done *within* POV-ray, before the resulting pixel
> ever reaches the monitor.)

The focal blur feature /is/ potentially problematic with respect to gamma; but
as it is performed well within POV-Ray's render engine, it is fine as long as
`assumed_gamma 1.0` is used.

As a matter of fact the potentially problematic operations aren't limited to
blurring, but encompass all operations where brightness values are added,
because the equation

    a = b + c

is /not/ equivalent to

    a^G = b^G + c^G

(except for a few special cases).

Blurring is problematic because it is an averaging operation, which in turn is
essentially an addition combined with a constant multiplication.

Post a reply to this message