Worley Cellular Basis Function

radiance-dev
1

Hi all,

Steven Worley has source code available for his cellular basis function. See www.texturingandmodeling.com

Is this something that could be incorporated as another function to call in procedural materials, similar to fnoise and others?

The copyright (see below) would seem to allow for use of the code in a pretty open manner.

Regards,

-Jack de Valpine

/* Copyright 1994, 2002 by Steven Worley
   This software may be modified and redistributed without restriction
   provided this comment header remains intact in the source code.
   This code is provided with no warrantee, express or implied, for
   any purpose.
      A detailed description and application examples can be found in the
   1996 SIGGRAPH paper "A Cellular Texture Basis Function" and
   especially in the 2002 book "Texturing and Modeling, a Procedural
   Approach, 3rd edition." There is also extra information on the web
   site http://www.worley.com/cellular.html .

   If you do find interesting uses for this tool, and especially if
   you enhance it, please drop me an email at steve@worley.com. */

/* Worley()

   An implementation of the key cellular texturing basis
   function. This function is hardwired to return an average F_1 value
   of 1.0. It returns the <n> most closest feature point distances
   F_1, F_2, .. F_n the vector delta to those points, and a 32 bit
   seed for each of the feature points. This function is not
   difficult to extend to compute alternative information such as
   higher order F values, to use the Manhattan distance metric, or
   other fun perversions.

   <at> The input sample location. <max_order> Smaller values compute faster. < 5, read the book to extend it.
   <F> The output values of F_1, F_2, ..F[n] in F[0], F[1], F[n-1]
   <delta> The output vector difference between the sample point and the n-th
            closest feature point. Thus, the feature point's location is the
            hit point minus this value. The DERIVATIVE of F is the unit
            normalized version of this vector.
   <ID> The output 32 bit ID number which labels the feature point. This
            is useful for domain partitions, especially for coloring flagstone
            patterns.

   This implementation is tuned for speed in a way that any order > 5
   will likely have discontinuous artifacts in its computation of F5+.
   This can be fixed by increasing the internal points-per-cube
   density in the source code, at the expense of slower
   computation. The book lists the details of this tuning. */

···

--
# John E. de Valpine
# president
#
# visarc incorporated
# http://www.visarc.com
#
# channeling technology for superior design and construction

2

Hi Jack,

I don't see why we couldn't link this in. I wonder how expensive it is -- I looked into implementing it myself following his Siggraph presentation, but it looked a bit too complex at the time. Taking the source code and adding it to the tree wouldn't be particularly difficult. I assume you'd find it useful if you're asking for it -- thanks for pointing out its availability.

-Greg

···

From: Jack de Valpine <jedev@visarc.com>
Date: July 8, 2004 1:38:03 PM PDT

Hi all,

Steven Worley has source code available for his cellular basis function. See www.texturingandmodeling.com

Is this something that could be incorporated as another function to call in procedural materials, similar to fnoise and others?

The copyright (see below) would seem to allow for use of the code in a pretty open manner.

Regards,

-Jack de Valpine

/* Copyright 1994, 2002 by Steven Worley
  This software may be modified and redistributed without restriction
  provided this comment header remains intact in the source code.
  This code is provided with no warrantee, express or implied, for
  any purpose.
    A detailed description and application examples can be found in the
  1996 SIGGRAPH paper "A Cellular Texture Basis Function" and
  especially in the 2002 book "Texturing and Modeling, a Procedural
  Approach, 3rd edition." There is also extra information on the web
  site http://www.worley.com/cellular.html .

  If you do find interesting uses for this tool, and especially if
  you enhance it, please drop me an email at steve@worley.com. */

/* Worley()

  An implementation of the key cellular texturing basis
  function. This function is hardwired to return an average F_1 value
  of 1.0. It returns the <n> most closest feature point distances
  F_1, F_2, .. F_n the vector delta to those points, and a 32 bit
  seed for each of the feature points. This function is not
  difficult to extend to compute alternative information such as
  higher order F values, to use the Manhattan distance metric, or
  other fun perversions.

  <at> The input sample location. <max_order> Smaller values compute faster. < 5, read the book to extend it.
  <F> The output values of F_1, F_2, ..F[n] in F[0], F[1], F[n-1]
  <delta> The output vector difference between the sample point and the n-th
           closest feature point. Thus, the feature point's location is the
           hit point minus this value. The DERIVATIVE of F is the unit
           normalized version of this vector.
  <ID> The output 32 bit ID number which labels the feature point. This
           is useful for domain partitions, especially for coloring flagstone
           patterns.

  This implementation is tuned for speed in a way that any order > 5
  will likely have discontinuous artifacts in its computation of F5+.
  This can be fixed by increasing the internal points-per-cube
  density in the source code, at the expense of slower
  computation. The book lists the details of this tuning. */