clipka <ano### [at] anonymousorg> wrote:
> Am 17.01.2013 12:14, schrieb Kenneth:
>
> >> If you'd use OpenEXR for the interim output, you could also get that
> >> bright light smear right.
> >
> > Really??! Then I need to do some research; the whole HDRI thing is kind of a
> > mystery to me, re: its use *in* computer graphics.
>
>
> It's really pretty simple, as far as POV-Ray is concerned:

See, that shows how much I *don't* know ;-)

I've been reading up on OpenEXR today--fascinating stuff (including the
odd/different gamma computations that it uses internally, to ultimately display
the image on typical 8-bit monitors. At least as a Photoshop plug-in. The
*display* of HDRI images is one of the things that I've never clearly
understood.) As a serendipitous side-effect, it has also launched me on a quest
to get a much better understanding of alpha compositing--specifically the
differences between using pre-multiplied alpha and non-pre-multiplied in an
image element.

For my animation here, I initially came up with a scheme to render the various
elements separately--meteors, moon and background--using POV-Ray's
Image_Alpha=on for the meteors and moon. Basically to see if I could save time
in the rendering process, as well as adding AA solely to the moon (6000 frames
for the meteors, but only 600 for the moon and background elements, as they
didn't need blurring, IMO.) I ran into some trouble though (in 3.62) when
averaging-together just the meteor elements as a pre-step (again using
Image_Alpha): The resulting meteor blur, when finally overlaid onto the other
elements, showed dark fringes. I now realize that alpha-multiplication (or non-)
is one of the culprits (as well as 3.7 having corrected some errors in how alpha
image_maps are applied.)

Knowledge is good! ;-)

So my animation here is from 6000 re-rendered 'normal' frames, no compositing.

But this scene might be the first (animation) experiment I re-render in 3.7RC6,
as separate elements again--this time using HDRI, assumed_gamma 1.0, and one or
the other of the alpha-multiplication schemes.

Thanks again for the OpenEXR suggestion.

Post a reply to this message