Top 
Functions
int  vips_conv () 
int  vips_convf () 
int  vips_convi () 
int  vips_conva () 
int  vips_convsep () 
int  vips_convasep () 
int  vips_compass () 
int  vips_gaussblur () 
int  vips_sharpen () 
int  vips_spcor () 
int  vips_fastcor () 
int  vips_sobel () 
int  vips_canny () 
Description
These operations convolve an image in some way, or are operations based on simple convolution, or are useful with convolution.
Functions
vips_conv ()
int vips_conv (VipsImage *in
,VipsImage **out
,VipsImage *mask
,...
);
Optional arguments:
precision
: VipsPrecision, calculation accuracylayers
:gint
, number of layers for approximationcluster
:gint
, cluster lines closer than this distance
Convolution.
Perform a convolution of in
with mask
.
Each output pixel is calculated as:
1 
sigma[i]{pixel[i] * mask[i]} / scale + offset 
where scale and offset are part of mask
.
By default, precision
is
VIPS_PRECISION_FLOAT. The output image
is always VIPS_FORMAT_FLOAT unless in
is VIPS_FORMAT_DOUBLE, in which case
out
is also VIPS_FORMAT_DOUBLE.
If precision
is VIPS_PRECISION_INTEGER, then
elements of mask
are converted to
integers before convolution, using rint()
,
and the output image
always has the same VipsBandFormat as the input image.
For VIPS_FORMAT_UCHAR images and VIPS_PRECISION_INTEGER precision
,
vips_conv()
uses a fast vector path based on
fixedpoint arithmetic. This can produce slightly different results.
Disable the vector path with vipsnovector
or VIPS_NOVECTOR
or
vips_vector_set_enabled()
.
If precision
is VIPS_PRECISION_APPROXIMATE then, like
VIPS_PRECISION_INTEGER, mask
is converted to int before convolution, and
the output image
always has the same VipsBandFormat as the input image.
Larger values for layers
give more accurate
results, but are slower. As layers
approaches the mask radius, the
accuracy will become close to exact convolution and the speed will drop to
match. For many large masks, such as Gaussian, n_layers
need be only 10% of
this value and accuracy will still be good.
Smaller values of cluster
will give more accurate results, but be slower
and use more memory. 10% of the mask radius is a good rule of thumb.
See also: vips_convsep()
.
[method]
Parameters
in 
input image 

out 
output image. 
[out] 
mask 
convolve with this mask 

... 

vips_convf ()
int vips_convf (VipsImage *in
,VipsImage **out
,VipsImage *mask
,...
);
Convolution. This is a lowlevel operation, see vips_conv()
for something
more convenient.
Perform a convolution of in
with mask
.
Each output pixel is
calculated as sigma[i]{pixel[i] * mask[i]} / scale + offset, where scale
and offset are part of mask
.
The convolution is performed with floatingpoint arithmetic. The output image
is always VIPS_FORMAT_FLOAT unless in
is VIPS_FORMAT_DOUBLE, in which case
out
is also VIPS_FORMAT_DOUBLE.
See also: vips_conv()
.
[method]
Parameters
in 
input image 

out 
output image. 
[out] 
mask 
convolve with this mask 

... 

vips_convi ()
int vips_convi (VipsImage *in
,VipsImage **out
,VipsImage *mask
,...
);
Integer convolution. This is a lowlevel operation, see vips_conv()
for
something more convenient.
mask
is converted to an integer mask with rint()
of each element, rint of
scale and rint of offset. Each output pixel is then calculated as
1 
sigma[i]{pixel[i] * mask[i]} / scale + offset 
The output image always has the same VipsBandFormat as the input image.
For VIPS_FORMAT_UCHAR images, vips_convi()
uses a fast vector path based on
halffloat arithmetic. This can produce slightly different results.
Disable the vector path with vipsnovector
or VIPS_NOVECTOR
or
vips_vector_set_enabled()
.
See also: vips_conv()
.
[method]
Parameters
in 
input image 

out 
output image. 
[out] 
mask 
convolve with this mask 

... 

vips_conva ()
int vips_conva (VipsImage *in
,VipsImage **out
,VipsImage *mask
,...
);
Optional arguments:
Perform an approximate integer convolution of in
with mask
.
This is a lowlevel operation, see
vips_conv()
for something more convenient.
The output image
always has the same VipsBandFormat as the input image.
Elements of mask
are converted to
integers before convolution.
Larger values for layers
give more accurate
results, but are slower. As layers
approaches the mask radius, the
accuracy will become close to exact convolution and the speed will drop to
match. For many large masks, such as Gaussian, layers
need be only 10% of
this value and accuracy will still be good.
Smaller values of cluster
will give more accurate results, but be slower
and use more memory. 10% of the mask radius is a good rule of thumb.
See also: vips_conv()
.
[method]
Parameters
in 
input image 

out 
output image. 
[out] 
mask 
convolution mask 

... 

vips_convsep ()
int vips_convsep (VipsImage *in
,VipsImage **out
,VipsImage *mask
,...
);
Optional arguments:
precision
: calculation accuracylayers
: number of layers for approximationcluster
: cluster lines closer than this distance
Perform a separable convolution of in
with mask
.
See vips_conv()
for a detailed description.
The mask must be 1xn or nx1 elements.
The image is convolved twice: once with mask
and then again with mask
rotated by 90 degrees. This is much faster for certain types of mask
(gaussian blur, for example) than doing a full 2D convolution.
See also: vips_conv()
, vips_gaussmat()
.
[method]
Parameters
in 
input image 

out 
output image. 
[out] 
mask 
convolution mask 

... 

vips_convasep ()
int vips_convasep (VipsImage *in
,VipsImage **out
,VipsImage *mask
,...
);
Optional arguments:
layers
:gint
, number of layers for approximation
Approximate separable integer convolution. This is a lowlevel operation, see
vips_convsep()
for something more convenient.
The image is convolved twice: once with mask
and then again with mask
rotated by 90 degrees.
mask
must be 1xn or nx1 elements.
Elements of mask
are converted to
integers before convolution.
Larger values for layers
give more accurate
results, but are slower. As layers
approaches the mask radius, the
accuracy will become close to exact convolution and the speed will drop to
match. For many large masks, such as Gaussian, layers
need be only 10% of
this value and accuracy will still be good.
The output image always has the same VipsBandFormat as the input image.
See also: vips_convsep()
.
[method]
Parameters
in 
input image 

out 
output image. 
[out] 
mask 
convolve with this mask 

... 

vips_compass ()
int vips_compass (VipsImage *in
,VipsImage **out
,VipsImage *mask
,...
);
Optional arguments:
times
:gint
, how many times to rotate and convolveangle
: VipsAngle45, rotate mask by this much between colvolutionscombine
: VipsCombine, combine results like thisprecision
: VipsPrecision, precision for blur, default floatlayers
:gint
, number of layers for approximationcluster
:gint
, cluster lines closer than this distance
This convolves in
with mask
times
times, rotating mask
by angle
each time. By default, it comvolves twice, rotating by 90 degrees, taking
the maximum result.
See also: vips_conv()
.
[method]
Parameters
in 
input image 

out 
output image. 
[out] 
mask 
convolve with this mask 

... 

vips_gaussblur ()
int vips_gaussblur (VipsImage *in
,VipsImage **out
,double sigma
,...
);
Optional arguments:
precision
: VipsPrecision, precision for blur, default intmin_ampl
: minimum amplitude, default 0.2
This operator runs vips_gaussmat()
and vips_convsep()
for you on an image.
Set min_ampl
smaller to generate a larger, more accurate mask. Set sigma
larger to make the blur more blurry.
See also: vips_gaussmat()
, vips_convsep()
.
[method]
Parameters
in 
input image 

out 
output image. 
[out] 
sigma 
how large a mask to use 

... 

vips_sharpen ()
int vips_sharpen (VipsImage *in
,VipsImage **out
,...
);
Optional arguments:
sigma
: sigma of gaussianx1
: flat/jaggy thresholdy2
: maximum amount of brighteningy3
: maximum amount of darkeningm1
: slope for flat areasm2
: slope for jaggy areas
Selectively sharpen the L channel of a LAB image. The input image is transformed to VIPS_INTERPRETATION_LABS.
The operation performs a gaussian blur and subtracts from in
to generate a
highfrequency signal. This signal is passed through a lookup table formed
from the five parameters and added back to in
.
The lookup table is formed like this:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 
. ^ . y2       .  / .  / slope m2 .  .../ . x1  ...  . ...> .  ...  x1 . ... slope m1 . /  . / m2  . /  . /  . /  . /  . ______/ _ _ _ _ _ _  y3 .  
For screen output, we suggest the following settings (the defaults):
1 2 3 4 5 6 
sigma == 0.5 x1 == 2 y2 == 10 (don't brighten by more than 10 L*) y3 == 20 (can darken by up to 20 L*) m1 == 0 (no sharpening in flat areas) m2 == 3 (some sharpening in jaggy areas) 
If you want more or less sharpening, we suggest you just change the m2 parameter.
The sigma
parameter changes the width of the fringe and can be
adjusted according to the output printing resolution. As an approximate
guideline, use 0.5 for 4 pixels/mm (display resolution),
1.0 for 12 pixels/mm and 1.5 for 16 pixels/mm (300 dpi == 12
pixels/mm). These figures refer to the image raster, not the halftone
resolution.
See also: vips_conv()
.
[method]
Parameters
in 
input image 

out 
output image. 
[out] 
... 

vips_spcor ()
int vips_spcor (VipsImage *in
,VipsImage *ref
,VipsImage **out
,...
);
Calculate a correlation surface.
ref
is placed at every position in in
and the correlation coefficient
calculated. The output
image is always float.
The output image is the same size as the input. Extra input edge pixels are made by copying the existing edges outwards.
The correlation coefficient is calculated as:
1 2 3 4 
sumij (ref(i,j)mean(ref))(inkl(i,j)mean(inkl)) c(k,l) =  sqrt(sumij (ref(i,j)mean(ref))^2) * sqrt(sumij (inkl(i,j)mean(inkl))^2) 
where inkl is the area of in
centred at position (k,l).
from Niblack "An Introduction to Digital Image Processing", Prentice/Hall, pp 138.
If the number of bands differs, one of the images must have one band. In this case, an nband image is formed from the oneband image by joining n copies of the oneband image together, and then the two nband images are operated upon.
The output image is always float, unless either of the two inputs is double, in which case the output is also double.
See also: vips_fastcor()
.
[method]
Parameters
in 
input image 

ref 
reference image 

out 
output image. 
[out] 
... 

vips_fastcor ()
int vips_fastcor (VipsImage *in
,VipsImage *ref
,VipsImage **out
,...
);
Calculate a fast correlation surface.
ref
is placed at every position in in
and the sum of squares of
differences calculated.
The output image is the same size as the input. Extra input edge pixels are made by copying the existing edges outwards.
If the number of bands differs, one of the images must have one band. In this case, an nband image is formed from the oneband image by joining n copies of the oneband image together, and then the two nband images are operated upon.
The output type is uint if both inputs are integer, float if both are float or complex, and double if either is double or double complex. In other words, the output type is just large enough to hold the whole range of possible values.
See also: vips_spcor()
.
[method]
Parameters
in 
input image 

ref 
reference image 

out 
output image. 
[out] 
... 

vips_sobel ()
int vips_sobel (VipsImage *in
,VipsImage **out
,...
);
Simple Sobel edge detector.
See also: vips_canny()
.
[method]
Parameters
in 
input image 

out 
output image. 
[out] 
... 

vips_canny ()
int vips_canny (VipsImage *in
,VipsImage **out
,...
);
Optional arguments:
sigma
:gdouble
, sigma for gaussian blurprecision
: VipsPrecision, calculation accuracy
Find edges by Canny's method: The maximum of the derivative of the gradient in the direction of the gradient. Output is float, except for uchar input, where output is uchar, and double input, where output is double. Noncomplex images only.
Use sigma
to control the scale over which gradient is measured. 1.4 is
usually a good value.
Use precision
to set the precision of edge detection. For uchar images,
setting this to VIPS_PRECISION_INTEGER will make edge detection much
faster, but sacrifice some sensitivity.
You will probably need to process the output further to eliminate weak edges.
See also: vips_sobel()
.
[method]
Parameters
in 
input image 

out 
output image. 
[out] 
... 
