I'm trying out some new techniques, and I thought I'd post a current render of
what I'm playing with.

It's just a technical render at this point, and so lacks any artistic merit
(unless someone has gone back in time and deleted every RSOCP image ever made).

One aspect is my own, very simplistic, HDR to light dome code. I've seen the
approaches people here have used in the past, but I needed something slightly
different. The basic algorithm is to simply over-sample the HDR image for
colours, and then choose the brightest N values. Choosing the brightest 100
samples from 10,000 samples can work very well for scenes with very bright, but
relatively small, lights.

Also, and it isn't showing up that well yet, I've got shaped focus blur to
emulate a proper lens iris. This one uses a pentagon shaped iris.

Finally the floor is a blurred reflector - which works very nicely except for a
few very bright lights giving sampled noise from the micro-facets. More samples
will fix the problem, but choosing a slightly less tricky light probe would hide
it too.

Cheers,
Edouard.

Post a reply to this message

Attachments:
Download 'wip-2008-07-26.jpg' (47 KB)

Preview of image 'wip-2008-07-26.jpg'

An earlier test render that used radiosity instead of a light dome. The light
dome gives much better and faster results for the HDR probe (but radiosity
could still be used usefully for inter-object illumination - Tek posted a scene
that mixed light domes and radiosity a year or more ago from memory).

The row of spheres on the left is a mirror finish, the second row is mildly
blurred (with dents as well), and the right row has a stronger blur (and deeper
dents). The floor is a blurry reflector as well.

The depth of field obviously needs more samples to hide the checkerboard moire.

Cheers,
Edouard.

Post a reply to this message

Attachments:
Download 'blur_by_blur.jpg' (35 KB)

Preview of image 'blur_by_blur.jpg'

I just realised, that, if I have thousands of samples of the light probe I'm
using in my scene, I can easily calculate a credible ambient light term.

Well, duh.

Post a reply to this message