| On this page |
float filterstep(float edge, float x, ...)
float filterstep(float edge, float x0, float x1, ...)
Returns the anti-aliased weight of the step function. The
step function returns 0 if x is less than edge and 1 if x is greater
than edge. filterstep returns a fraction representing the
filtered area under the step function. Filtering is computed using
derivatives Du() and Dv() which are only non-zero in shading and COP
contexts, so filterstep will not perform filtering in other contexts.
Image filtering options ¶
Examples of specifying filter parameters:
colormap(map, u, v, "smode", "decal", "tmode", "repeat", "border", {.1,1,1}); colormap(map, u, v, "mode", "clamp", "width", 1.3); colormap(map, u, v, "filter", "gauss", "width", 1.3, "mode", "repeat");
If the texture is a deep .rat file, you can use the "channel" keyword argument
to specify a channel in the file:
string channelname = "N"; cf = colormap(map, u, v, "channel", channelname);
-
When you read a texture in a format other than Houdini’s native
.picor.rat, Houdini uses OpenImageIO to read the image data from the file. In that case, some of the variadic arguments below may not have any effect. -
When the texture function evaluates non-houdini format textures, Houdini switches to use OpenImageIO for texture evaluation. While there are corresponding values to many of the variadic keywords, some keywords don’t have an equivalent function in OpenImageIO.
-
OIIO will not create MIP maps for images that don’t have multi-resolution images by default. You can turn this on by adding
automip=1to the content of the theOPENIMAGEIO_IMAGECACHE_OPTIONSenvironment variable.Without MIP maps, blurring and filtering may not work as expected.
-
You can also use
OPENIMAGEIO_IMAGECACHE_OPTIONSto override the amount of memory OIIO uses for caching.By default, Houdini will set the cache memory to 1/8th of the physical computer memory. If you set the
OPENIMAGEIO_IMAGECACHE_OPTIONSvariable it overrides that computed cache size.
-
"wrap",
string
="repeat"
repeat or periodic
The image map will repeat outside the range 0 to 1. Basically, the integer component of the texture coordinate is ignored. This is the default.
clamp or edge or streak
The texture coordinates will be clamped to the range 0 to 1. This causes evaluations outside the range to evaluate to the color at the closest edge of the image (the border pixels are streaked outside the range).
black or decal or color
Coordinates outside the range 0 to 1 will evaluate to the border color (rather than a color in the image). The border color is black (i.e. 0) by default.
"uwrap",
string
(AKA swrap)
Specifies the behavior when the u coordinate is outside
the range 0 to 1. The values are the same as with wrap.
"vwrap",
string
(AKA twrap)
Specifies the behavior when the v coordinate is outside
the range 0 to 1. The values are the same as with wrap.
"border",
float|vector|vector4
=0
Specifies the border color when Black/Decal/Color wrapping is used. Has no effect for OpenImageIO formats.
"default_color",
float|vector|vector4
Specifies the color to use when the texture map cannot be found. If this argument is not given, the color is set by the HOUDINI_DEFAULT_TEXTURE_COLOR variable.
"channel",
Specifies the color channel for textures that have multiple color
planes (for example, diffuse_indirect or N).
For ptex images, this specifies the index of the first channel
(for example, 0 or 4).
"blur",
float
Blurs in x and y directions. Blur is measured as a percentage
of the image size - so a blur of 0.1 will blur 10% of the image
width. Use xblur and yblur if you need different blur
amounts in either dimension.
"xblur",
(AKA ublur, sblur)
Blur amount in the x image direction.
"yblur",
(AKA vblur, tblur)
Blur amount in the y image direction.
"pixelblur",
float
Blurs the texture by a floating point number of pixels. Has no effect for OpenImageIO formats.
Cf = texture("map.rat", ss, tt, "pixelblur", 2.0);
"xpixelblur",
float
Blurs the texture by a floating point number of pixels in the X direction.
"ypixelblur",
float
Blurs the texture by a floating point number of pixels in the Y direction.
"filter",
string
="box"
Specifies the type of anti-aliasing filter to be used for evaluation.
For Houdini native formats, the following value should be a string specifying one of:
"point"
Point sampling (i.e. no filtering)
"box"
Box filter (default)
"gauss"
Gaussian filter
"bartlett"
Bartlett/Triangular filter
"sinc"
Sinc sharpening filter
"hanning"
Hanning filter
"blackman"
Blackman filter
"catrom"
Catmull-Rom filter
For all other formats (loaded by OpenImageIO), specifying the "point" filter sets the OIIO interpolation mode to "closest" and disables MIP mapping. Any other value uses OIIO smart-bicubic interpolation. You can get finer control using the "filtermode" variadic argument (see below).
"xfilter",
string
(AKA ufilter, sfilter)
Specifies the filter for the X direction. The filters are
the same as with filter.
"yfilter",
string
(AKA vfilter, tfilter)
Specifies the filter for the Y direction. The filters are
the same as with filter.
"filtermode",
string
For Houdini native formats, VEX also supports simpler filtering. The
filtermode can be set to one of:
filter
Use the filter specified by the filter keyword argument.
bilinear
Use simple bilinear filtering. This is the fastest specialized filtering mode, but provides the lowest quality filtering.
biquadratic
Use simple quadratic filtering (order 3 filtering).
bicubic
Use simple bicubic filtering.
When the filtermode is set to bilinear, biquadratic or bicubic,
several arguments (like filter and width) are ignored and a fixed
interpolation filter is used instead. Other arguments (notably the lerp
and blur keywords) are still valid.
For all other formats (loaded by OpenImageIO) you can set the filtermode to "filter" (see "filter" above), "bilinear", "biquadratic", or "bicubic".
"width",
float
=1.0
For Houdini native formats, this sets the filter width in both X and Y directions.
For all other formats (loaded by OpenImageIO), this sets the OIIO swidth and twidth options.
"xwidth",
float
(AKA uwidth, swidth)
Filter width in the X direction.
"ywidth",
float
(AKA vwidth, twidth)
Filter width in the Y direction.
"zwidth",
float
Filter width in the Z direction (for shadow maps). This is measured in world space units, unlike the other width arguments.
"extrapolate",
int
whether to use derivative extrapolation when computing anti-aliasing information. Extrapolation of derivatives is on by default. The argument should be either 0 or 1.
"lerp",
int
For Houdini native formats, this specifies whether RAT files should interpolate between different MIP levels. By default, this is turned off. Turning interpolation on will help remove discontinuities when different
MIP levels of a .rat file are accessed. However, the results of texture evaluation will be slightly softer (i.e. blurrier) and will take more time.
There are three possible values for this argument.
0
Disable MIP map interpolation (fastest).
1
Approximate MIP map interpolation (fast).
2
High Quality MIP map interpolation (slower but highest quality).
For all other formats (loaded by OpenImageIO), a value of 0 specifies a single MIP level, any other value specifies trilinear interpolation.
"depthinterp",
string
Specifies the depth interpolation mode for deep shadow maps, to control the opacity value that will be returned when the map is sampled between two z-records.
The argument must be a string.
discrete
(default) Return the first z-record before the sample point.
linear
Linearly interpolate the opacities of the z-records before and after the sample point.
See deep shadow maps for more on the difference between the two modes.
"beerlambert",
int
When evaluating volumetric deep shadow maps, this will enable Beer-Lambert interpolation of opacity. Beer-Lambert is more a accurate but more expensive form of interpolation.
The argument should be either 0 or 1.
"srccolorspace",
string
Specifies the color space in which the texture is stored. When texture values are accessed, they will be translated from this space into linear space for rendering if needed.
auto
(default) Determine the source color space based on the file. Currently, this will assume sRGB color space for 8-bit textures and linear for all other textures.
linear
Transform to linear space. This currently only affects 8-bit textures, since all others are assumed to be already in linear space. Use this option to force linear interpretation of textures used for bump or displacement maps.
sRGB
Forcibly translate from sRGB color space to linear space regardless of the bit-depth or number of channels in the the texture.
rec709
Convert from Rec709 color space to linear space.
gamma22
Convert from Gamma 2.2 color space to linear space.
raw
Use map colors untransformed
The srccolorspace argument can also be any color space known to OpenColorIO.
"face",
When using a Ptex texture map, the face argument is used to specify the face for ptexture lookup.
Has no effect for OpenImageIO formats.
"ptexorient",
int
When using Ptex textures, the implicit texture coordinates on
polygons are used as the interpolants for texture lookup (combined
with the face). However, different software may have different
beliefs about winding and orientation. This keyword argument
allows you to control the interpretation of orientation for Houdini
polygons. The ptexorient expects an integer argument which is
composed of a bit-field
-
bit 0×01: Complement the
scoordinate -
bit 0×02: Complement the
tcoordinate -
bit 0×04: Swap the
sandtcoordinates
For example, a value of 6 (0×4|0×2) is equivalent to calling
texture(map, 1-t, s) instead of texture(map, s, t).
The default ptexorient is 0, which works correctly with the
examples found at http://ptex.us.
Has no effect for OpenImageIO formats.
"iesnormalization",
string
="maxvalue"
Select different methods of normalizing IES map’s output values when
querying via environment() function.
none
Use raw values scaled by the candela multiplier in the header.
maxvalue
(default) Normalized by the maximum value. This is legacy behavior used by mantra’s default light shader.
preserveenergy
Normalized by values integrated over coverage angles, so that IES profile affects shaping of the light while preserving its overall energy output.
Examples ¶
f = filterstep(0.5, s+t, "filter", "gauss", "width", 2);
The filterstep(float edge, x, ...) form is roughly equivalent to:
f = filterstep(edge, x, x + abs(Du(x) + Dv(x)));
| shading |
|