| 1 |
.\" RCSid $Id: bsdf2ttree.1,v 1.12 2023/09/12 16:40:57 greg Exp $ |
| 2 |
.TH BSDF2TTREE 1 4/24/2013 RADIANCE |
| 3 |
.SH NAME |
| 4 |
bsdf2ttree - generate XML tensor tree description of a BSDF |
| 5 |
.SH SYNOPSIS |
| 6 |
.B bsdf2ttree |
| 7 |
[ |
| 8 |
.B "\-pC" |
| 9 |
][ |
| 10 |
.B "{+|-}a" |
| 11 |
][ |
| 12 |
.B "\-g Nlog2" |
| 13 |
][ |
| 14 |
.B "\-t pctcull" |
| 15 |
][ |
| 16 |
.B "\-n nss" |
| 17 |
][ |
| 18 |
.B "\-s thresh" |
| 19 |
][ |
| 20 |
.B "\-l maxlobes" |
| 21 |
][ |
| 22 |
.B "bsdf.sir .." |
| 23 |
] |
| 24 |
.br |
| 25 |
or |
| 26 |
.br |
| 27 |
.B bsdf2ttree |
| 28 |
.B "\-t{3|4}" |
| 29 |
[ |
| 30 |
.B "\-pC" |
| 31 |
][ |
| 32 |
.B "{+|-}a" |
| 33 |
][ |
| 34 |
.B "\-g Nlog2" |
| 35 |
][ |
| 36 |
.B "\-t pctcull" |
| 37 |
][ |
| 38 |
.B "\-n nss" |
| 39 |
][ |
| 40 |
.B "\-s thresh" |
| 41 |
][ |
| 42 |
.B "{+|-}forward" |
| 43 |
][ |
| 44 |
.B "{+|-}backward" |
| 45 |
][ |
| 46 |
.B "\-e expr" |
| 47 |
][ |
| 48 |
.B "\-f file" |
| 49 |
] |
| 50 |
.B bsdf_func |
| 51 |
.SH DESCRIPTION |
| 52 |
.I Bsdf2ttree |
| 53 |
produces a tensor tree representation of a |
| 54 |
bidirectional scattering distribution function (BSDF) |
| 55 |
based on an intermediate representation (in the first form) or |
| 56 |
a functional description (in the second form). |
| 57 |
A complete XML description is written to the standard output, |
| 58 |
which is normally redirected to a file. |
| 59 |
.PP |
| 60 |
The |
| 61 |
.I \-p |
| 62 |
option by itself turns off the progress bar, whose length may be set |
| 63 |
by an immediately following integer argument. |
| 64 |
(The default progress bar length is 79 characters.)\0 |
| 65 |
.PP |
| 66 |
The |
| 67 |
.I \+a |
| 68 |
option turns on reciprocity averaging for isotropic scattering or anisotropic reflection. |
| 69 |
Normally off (-a), this enforces each tensor-tree hemisphere to follow symmetry |
| 70 |
implied by Helmholtz reciprocity, and is designed to reduce noise in measured data. |
| 71 |
However, for some data, reciprocity averaging can make the output worse. |
| 72 |
.PP |
| 73 |
The maximum resolution of the tensor tree may be controlled by the |
| 74 |
.I \-g |
| 75 |
option, which defaults to a value of 6. |
| 76 |
This corresponds to a peak resolution of 2^6 (64) in each dimension. |
| 77 |
Due to memory and time constraints, it is not recommended to set |
| 78 |
.I \-g |
| 79 |
higher than 7, which corresponds to a 128x128x128x128 initial sampling, |
| 80 |
or 268 million values. |
| 81 |
.PP |
| 82 |
The initial sampling is pared down by the percentage specified with the |
| 83 |
.I \-t |
| 84 |
option, which defaults to 90. |
| 85 |
Setting this parameter to -1 turns culling off entirely, which may be |
| 86 |
useful for comparisons. |
| 87 |
.PP |
| 88 |
The |
| 89 |
.I \-n |
| 90 |
option controls the number of super-samples to send in patches whose |
| 91 |
difference to its neighbors exceeds some threshold. |
| 92 |
The default number of super-samples is 64. |
| 93 |
The difference threshold for super-sampling is controlled by the |
| 94 |
.I \-s |
| 95 |
option, and defaults to 0.35. |
| 96 |
.PP |
| 97 |
The first invocation form takes a intermediate scattering representation |
| 98 |
as produced by |
| 99 |
.I pabopto2bsdf(1) |
| 100 |
or similar, and produces a tensor tree representation with as many |
| 101 |
components as there are independent input distributions. |
| 102 |
Each intermediate scattering file contains one of |
| 103 |
the four components, and if the first component |
| 104 |
is isotropic, all components must be isotropic. |
| 105 |
A similar rule holds for anisotropic inputs. |
| 106 |
The |
| 107 |
.I \-l |
| 108 |
option may be used to specify the maximum number of lobes in any |
| 109 |
interpolated radial basis function. |
| 110 |
The default value is 15000, which generally keeps the interpolation tractable. |
| 111 |
Setting the value to 0 turns off this limit. |
| 112 |
Parameter options may be altered between input files, in case a different |
| 113 |
resolution or culling percentage is indicated for transmission versus |
| 114 |
reflection for example. |
| 115 |
.PP |
| 116 |
In the second invocation form, |
| 117 |
.I bsdf2ttree |
| 118 |
takes a functional specification of a BSDF. |
| 119 |
The named function should accept 6 parameters corresponding to the |
| 120 |
normalized incident and exiting vectors, respectively. |
| 121 |
By convention, these vectors point away from the surface, and a positive |
| 122 |
Z-component corresponds to the front side. |
| 123 |
The Y-component corresponds to the "up" orientation of the surface, |
| 124 |
as specified in the eventual scene description that references the XML |
| 125 |
output. |
| 126 |
If the function only takes 3 parameters, then the variables "Dx", "Dy", |
| 127 |
and "Dz" will be assigned to the reverse of the outgoing direction at |
| 128 |
each evaluation. |
| 129 |
(I.e., the vector will point into the surface and |
| 130 |
Dz will be negative on the front side.)\0 |
| 131 |
This simplifies conversion of functional BSDF specifications using the |
| 132 |
legacy material primitives "plasfunc", "metfunc", and "transfunc". |
| 133 |
.PP |
| 134 |
The function is defined by one or more |
| 135 |
.I \-e |
| 136 |
and |
| 137 |
.I \-f |
| 138 |
options, and should obey both Helmholtz reciprocity and |
| 139 |
integrate to less than 1 over each projected incident hemisphere |
| 140 |
for energy conservation. |
| 141 |
The variable and function definitions in each |
| 142 |
.I \-f source |
| 143 |
file are read and compiled from the RADIANCE library where it is found. |
| 144 |
If the |
| 145 |
.I \-t3 |
| 146 |
option is specified, the defined function is assumed to be isotropic. |
| 147 |
If the |
| 148 |
.I \-t4 |
| 149 |
option is given, the function is assumed to be anisotropic. |
| 150 |
.PP |
| 151 |
Similar to the |
| 152 |
.I genBSDF(1) |
| 153 |
command, |
| 154 |
the |
| 155 |
.I \+backward |
| 156 |
option (default) specifies that rays arriving from the front side of |
| 157 |
the surface will be tested for reflection and transmission. |
| 158 |
If both forward and backward (front and back) distributions are needed, the |
| 159 |
.I \+forward |
| 160 |
option may be given. |
| 161 |
To turn off the backward components, use the |
| 162 |
.I \-backward |
| 163 |
option. |
| 164 |
Computing both incident hemispheres takes about twice as long as one, but |
| 165 |
is recommended when rays will be impinging from either side. |
| 166 |
.SH EXAMPLE |
| 167 |
To take two components of an intermediate BSDF representation and create |
| 168 |
a high-resolution tensor tree with 85% culling on transmission and 95% |
| 169 |
culling on reflection: |
| 170 |
.IP "" .2i |
| 171 |
bsdf2ttree -g 7 -t 85 transmitted.sir -t 95 reflected.sir > combined.xml |
| 172 |
.PP |
| 173 |
To create a low-res BSDF corresponding to a one-sided, |
| 174 |
isotropic Phong distribution: |
| 175 |
.IP "" .2i |
| 176 |
bsdf2ttree -g 5 -t3 -e 'phong(ix,iy,iz,ox,oy,oz) = if(iz, .1+((iz+oz)/sqrt((ix+ox)^2+(iy+oy)^2+(iz+oz)^2))^50, 0)' phong > phong.xml |
| 177 |
.SH ENVIRONMENT |
| 178 |
RAYPATH the directories to check for auxiliary files. |
| 179 |
.SH AUTHOR |
| 180 |
Greg Ward |
| 181 |
.SH "SEE ALSO" |
| 182 |
bsdf2klems(1), icalc(1), genBSDF(1), pabopto2bsdf(1), pabopto2xyz(1), |
| 183 |
pkgBSDF(1), rcontrib(1), rfluxmtx(1), wrapBSDF(1) |