I finally found the time to add automatic image comparison to my POV-Ray 
version benchmarking environment, to spit out stuff like this now:

The "optics.pov" sample scene rendered with POV-Ray 3.6.1 (top left) and 
an experimental 3.7.0 variant of mine (virtually identical to beta.34 
for the sake of this image); bottom left: The "delta" of the two images 
(exaggerated for clarity, and somewhat blurred to reduce the sensitivity 
to differences caused by jitter); bottom right: A b/w version of the 
image, with differing areas "markered" in green (extremely minor 
differences, <= 1%), yellow (minor differences, <= 10%) and red (major 
differences, > 10%). The program will also give feedback via return code 
how severe the differences are.

Comparison is performed in "gamma space" using a gamma of 2.2, as I'm 
interested in perceptual differences, not logical ones.

(No, unfortunately it's not a POV-Ray script that does it ;-) It's 
actually a small C program thrown together within about two and a half 
hours from draft to what I consider finalized for now; my first C 
program to use libpng, besides.)

Post a reply to this message

Attachments:
Download 'optics.pov.png' (1585 KB)

Preview of image 'optics.pov.png'

Ah, the beauty of errors... (see bottom left).

This is the "newdiffract.pov" sample scene, rendered with POV-Ray 3.6 
vs. POV-Ray 3.7.0.beta.29 (but beta.34 renders virtually identical).

BTW, the difference in hue is apparently due to the problems with 
dispersion on which I posted recently; with the patch I created to fix 
those issues, the scene renders much more like 3.6 again.

Post a reply to this message

Attachments:
Download 'newdiffract.pov.png' (1959 KB)

Preview of image 'newdiffract.pov.png'