but let's go back to the interesting observation that jr made
Too few polygons in some places (it was a car signal light) are not nice
You're right, I thought about it.

General algorithms are no good in my opinion.
They are recursive and have unintelligent ways.
They consume memory.
They exaggerate in the points that are already rich and perhaps are deficient in
the points that need it most.

I thought of a different criterion, to be applied only on "poor" polygons.
It is not easy but not impossible to write.
I attach the scheme of my idea.
I have 3 quads (seen from above).

I select the central one B, to improve it by splitting it in two.

Above, the 3 black quads with their normals red.
In the center, their green vertex normals, computed for POVray's smooth
triangles.


Below, I've broken B into two parts, B and B1. Finding the split point as the
midpoint between its original midline and the intersection of its lateral quads.


Now I have to recalculate the new normals ...
blue and yellow.
The bad quad is gone. If it's not enough I can select another bad quad.
Here I simplified, the sides are symmetrical, but it's just to simplify the
drawing. You can always do it.

When I have two free hours to concentrate I try.

Post a reply to this message

Attachments:
Download 'clipboard01.png' (6 KB)

Preview of image 'clipboard01.png'