Elsewhere the idea of leaning toward a point in space as a normal 
pattern was discussed. In an initial povr branch implementation ended up 
with two forms of the pattern. One for use in averages with other 
patterns and another which comes after the zero weighted perturbed 
pattern prior that adds the lean on top of the previous perturbation. 
These are 'bevy type 3' and 'bevy type 4' below, respectively.

In the attached image the upper left corner is the original sphere with 
no normal{} perturbation. In the upper right only a lean toward +y. In 
the lower left an average of normals where one pattern wrinkles like and 
the other the lean toward point. In the lower right the second form 
which leans the already wrinkle perturbed normal to create a final 
perturbed normal.

The code for the lower two images looks like:

  ...
  average
  normal_map {
       [1 bevy type 1 bump_size 0.5]
       [1 bevy type 3 bump_size 0.5] // lean
  }
  ...

and

  ...
  average
  normal_map {
       [0 bevy type 1 bump_size 0.5]
       [1 bevy type 4 bump_size 0.5] // lean normal perturbed above
  }
  ...

where currently 'bevy' is a normal perturbation pattern really a 
collection of experimental normal patterns of which two are the leaning 
kind(1). Type 1 is a possible non-biased wrinkles replacement.

Bill P.

(1) - Wondering if types not the better way to go over keywords when 
patterns are related - otherwise the number of keywords starts to explode.

Post a reply to this message

Attachments:
Download 'normalleanstory.jpg' (24 KB)

Preview of image 'normalleanstory.jpg'