On 6/17/2012 8:53 AM, clipka wrote:
> Am 17.06.2012 13:19, schrieb FlyerX:
>> Hello,
>>
>> I am working with the SSLT features in RC5 (Windows 64bit) and so far
>> they seem to give nice results and I am adding it to PoseRay as a
>> material property.
>>
>> According to the POV-Ray documentation, there is additional information
>> about the translucency for several materials listed in
>> http://graphics.stanford.edu/papers/bssrdf/bssrdf.pdf.
>>
>> I suppose the documentation refers to Figure 5 in the SIGGRAPH
>> publication. Which of the values (or combination of them) are we
>> supposed to use? I was given by clipka a skin value of <2.0,1.5,0.8>
>> which, according to him, should be between the skin1 and skin2 values in
>> Figure 5 but does not seem to match the data.
>
> The SSLT parameterization has been changed since its introduction, to
> allow for more intuitive tweaking, and also for easier use of patterned
> pigments.
>
> The early implementation was parameterized by the "reduced scattering
> coefficient" (sigma-prime[s]) and the "absorption coefficient"
> (sigma[alpha]); as these two parameters completely define a material's
> apparent color, the materials regular pigment was ignored; finding
> suitable parameters for materials not listed in the paper was
> cumbersome, as the relation between the coefficients and the resulting
> color is far from intuitive. Furthermore, using image maps to e.g. add
> some variation to a skin texture would have been impossible with this
> approach.
>
> The current implementation takes two different parameters, from which it
> then computes the aforementioned coefficients internally. These
> parameters are the material's "diffuse reflectance" (R[d]) - i.e. the
> resulting apparent color of the material, which is conveniently
> specified via the classic pigment statement (modulated by the diffuse
> parameter) - and the "mean free path" (the inverse of sigma[tr]) - which
> is what you specify via the "translucency" parameter.
>
>
> This parameterization has been suggested by Jensen & Buhler in a
> follow-up paper, "A Rapid Hierarchical Rendering Technique for
> Translucent Materials", but they didn't compute the corresponding values
> for the materials listed in the earlier paper.
>
> Of course, from the equations in the two papers a formula can be thrown
> together to compute the translucency parameter from the measured values,
> but so far I haven't found the time and patience.

Thanks for the clarification. In that case I went ahead and calculated 
the translucency or the inverse of the Effective extinction coefficient 
and put the values below:

Apple		6.96	6.40	1.90
Chicken1	11.61	3.88	1.75
Chicken2	9.44	3.35	1.79
Cream		15.03	4.66	2.54
Ketchup		4.76	0.57	0.39
Marble		8.51	5.57	3.95
Potato		14.27	7.23	2.04
Skimmilk	18.42	10.44	3.50
Skin1		3.67	1.37	0.68
Skin2		4.82	1.69	1.09
Wholemilk	10.90	6.58	2.51

The skin values do not match what you gave me but I may have made a 
calculation error. The relative magnitude of the values seems to make 
sense. Larger ones for skim milk and low ones for skin and apple. I 
calculated the values as follows:

sigma_t(prime)=sigma_s(prime)+sigma_a,  (both are given in Figure 5)
sigma_tr=sqrt[3 * sigma_a * sigma_t(prime) ]
translucency = mean free path = 1/sigma_tr

the math was done on each RGB component separately.

later,

FlyerX

Post a reply to this message