resample

resample — resample images in various ways

Stability Level

Stable, unless otherwise indicated

Functions

int vips_shrink ()
int vips_shrinkh ()
int vips_shrinkv ()
int vips_reduce ()
int vips_reduceh ()
int vips_reducev ()
int vips_thumbnail ()
int vips_thumbnail_buffer ()
int vips_similarity ()
int vips_affine ()
int vips_resize ()
int vips_mapim ()
int vips_quadratic ()

Types and Values

enum VipsKernel
enum VipsSize

Includes

#include <vips/vips.h>

Description

There are three types of operation in this section.

First, vips_affine() applies an affine transform to an image. This is any sort of 2D transform which preserves straight lines; so any combination of stretch, sheer, rotate and translate. You supply an interpolator for it to use to generate pixels, see vips_interpolate_new(). It will not produce good results for very large shrinks: you'll see aliasing.

vips_reduce() is like vips_affine(), but it can only shrink images, it can't enlarge, rotate, or skew. It's very fast and uses an adaptive kernel for interpolation. Again, it will give poor results for large size reductions.

vips_shrink() is a fast block shrinker. It can quickly reduce images by large integer factors. It will give poor results for small size reductions: again, you'll see aliasing.

Next, vips_resize() specialises in the common task of image reduce and enlarge. It strings together combinations of vips_shrink(), vips_reduce(), vips_affine() and others to implement a general, high-quality image resizer.

Finally, vips_mapim() can apply arbitrary 2D image transforms to an image.

Functions

vips_shrink ()

int
vips_shrink (VipsImage *in,
             VipsImage **out,
             double hshrink,
             double vshrink,
             ...);

Shrink in by a pair of factors with a simple box filter. For non-integer factors, vips_shrink() will first shrink by the integer part with a box filter, then use vips_reduce() to shrink by the remaining fractional part.

This is a very low-level operation: see vips_resize() for a more convenient way to resize images.

This operation does not change xres or yres. The image resolution needs to be updated by the application.

See also: vips_resize(), vips_reduce().

Parameters

in

input image

 

out

output image

 

hshrink

horizontal shrink

 

vshrink

vertical shrink

 

...

NULL-terminated list of optional named arguments

 

Returns

0 on success, -1 on error


vips_shrinkh ()

int
vips_shrinkh (VipsImage *in,
              VipsImage **out,
              int hshrink,
              ...);

Shrink in horizontally by an integer factor. Each pixel in the output is the average of the corresponding line of hshrink pixels in the input.

This is a very low-level operation: see vips_resize() for a more convenient way to resize images.

This operation does not change xres or yres. The image resolution needs to be updated by the application.

See also: vips_shrinkv(), vips_shrink(), vips_resize(), vips_affine().

Parameters

in

input image

 

out

output image

 

hshrink

horizontal shrink

 

...

NULL-terminated list of optional named arguments

 

Returns

0 on success, -1 on error


vips_shrinkv ()

int
vips_shrinkv (VipsImage *in,
              VipsImage **out,
              int vshrink,
              ...);

Shrink in vertically by an integer factor. Each pixel in the output is the average of the corresponding column of vshrink pixels in the input.

This is a very low-level operation: see vips_resize() for a more convenient way to resize images.

This operation does not change xres or yres. The image resolution needs to be updated by the application.

See also: vips_shrinkh(), vips_shrink(), vips_resize(), vips_affine().

Parameters

in

input image

 

out

output image

 

vshrink

vertical shrink

 

...

NULL-terminated list of optional named arguments

 

Returns

0 on success, -1 on error


vips_reduce ()

int
vips_reduce (VipsImage *in,
             VipsImage **out,
             double hshrink,
             double vshrink,
             ...);

Optional arguments:

  • kernel : VipsKernel to use to interpolate (default: lanczos3)

  • centre : gboolean use centre rather than corner sampling convention

Reduce in by a pair of factors with a pair of 1D kernels. This will not work well for shrink factors greater than three.

Set centre to use centre rather than corner sampling convention. Centre convention can be useful to match the behaviour of other systems.

This is a very low-level operation: see vips_resize() for a more convenient way to resize images.

This operation does not change xres or yres. The image resolution needs to be updated by the application.

See also: vips_resize(), vips_affine().

Parameters

in

input image

 

out

output image

 

hshrink

horizontal shrink

 

vshrink

vertical shrink

 

...

NULL-terminated list of optional named arguments

 

Returns

0 on success, -1 on error


vips_reduceh ()

int
vips_reduceh (VipsImage *in,
              VipsImage **out,
              double hshrink,
              ...);

vips_reducev ()

int
vips_reducev (VipsImage *in,
              VipsImage **out,
              double vshrink,
              ...);

vips_thumbnail ()

int
vips_thumbnail (const char *filename,
                VipsImage **out,
                int width,
                ...);

Optional arguments:

  • height : gint, target height in pixels

  • size : VipsSize, upsize, downsize or both

  • auto_rotate : gboolean, rotate upright using orientation tag

  • crop : VipsInteresting, shrink and crop to fill target

  • linear : gboolean, perform shrink in linear light

  • import_profile : gchararray, fallback import ICC profile

  • export_profile : gchararray, export ICC profile

Make a thumbnail from a file. Shrinking is done in three stages: using any shrink-on-load features available in the file import library, using a block shrink, and using a lanczos3 shrink. At least the final 200% is done with lanczos3. The output should be high quality, and the operation should be quick.

See vips_thumbnail_buffer() to thumbnail from a memory source.

The output image will fit within a square of size width x width . You can specify a separate height with the height option.

If you set crop , then the output image will fill the whole of the width x height rectangle, with any excess cropped away. See vips_smartcrop() for details on the cropping strategy.

Normally the operation will upsize or downsize as required. If size is set to VIPS_SIZE_UP, the operation will only upsize and will just copy if asked to downsize. If size is set to VIPS_SIZE_DOWN, the operation will only downsize and will just copy if asked to upsize.

Normally any orientation tags on the input image (such as EXIF tags) are interpreted to rotate the image upright. If you set auto_rotate to FALSE, these tags will not be interpreted.

Shrinking is normally done in sRGB colourspace. Set linear to shrink in linear light colourspace instead --- this can give better results, but can also be far slower, since tricks like JPEG shrink-on-load cannot be used in linear space.

If you set export_profile to the filename of an ICC profile, the image will be transformed to the target colourspace before writing to the output. You can also give an import_profile which will be used if the input image has no ICC profile, or if the profile embedded in the input image is broken.

See also: vips_thumbnail_buffer().

Parameters

filename

file to read from

 

out

output image

 

width

target width in pixels

 

...

NULL-terminated list of optional named arguments

 

Returns

0 on success, -1 on error.


vips_thumbnail_buffer ()

int
vips_thumbnail_buffer (void *buf,
                       size_t len,
                       VipsImage **out,
                       int width,
                       ...);

Optional arguments:

  • height : gint, target height in pixels

  • size : VipsSize, upsize, downsize or both

  • auto_rotate : gboolean, rotate upright using orientation tag

  • crop : VipsInteresting, shrink and crop to fill target

  • linear : gboolean, perform shrink in linear light

  • import_probuffer : gchararray, fallback import ICC probuffer

  • export_probuffer : gchararray, export ICC probuffer

Exacty as vips_thumbnail(), but read from a memory buffer.

See also: vips_thumbnail().

Parameters

buf

memory area to load

 

len

size of memory area

 

out

output image

 

width

target width in pixels

 

...

NULL-terminated list of optional named arguments

 

Returns

0 on success, -1 on error.


vips_similarity ()

int
vips_similarity (VipsImage *in,
                 VipsImage **out,
                 ...);

Optional arguments:

  • scale : gdouble, scale by this factor

  • angle : gdouble, rotate by this many degrees clockwise

  • interpolate : VipsInterpolate, interpolate pixels with this

  • idx : gdouble, input horizontal offset

  • idy : gdouble, input vertical offset

  • odx : gdouble, output horizontal offset

  • ody : gdouble, output vertical offset

This operator calls vips_affine() for you, calculating the matrix for the affine transform from scale and angle . Other parameters are passed on to vips_affine() unaltered.

See also: vips_affine(), VipsInterpolate.

Parameters

in

input image

 

out

output image

 

...

NULL-terminated list of optional named arguments

 

Returns

0 on success, -1 on error


vips_affine ()

int
vips_affine (VipsImage *in,
             VipsImage **out,
             double a,
             double b,
             double c,
             double d,
             ...);

Optional arguments:

This operator performs an affine transform on an image using interpolate .

The transform is:

X = a * (x + idx ) + b * (y + idy ) + odx Y = c * (x + idx ) + d * (y + idy ) + doy

x and y are the coordinates in input image. X and Y are the coordinates in output image. (0,0) is the upper left corner.

The section of the output space defined by oarea is written to out . oarea is a four-element int array of left, top, width, height. By default oarea is just large enough to cover the whole of the transformed input image.

interpolate defaults to bilinear.

idx , idy , odx , ody default to zero.

This operation does not change xres or yres. The image resolution needs to be updated by the application.

See also: vips_shrink(), vips_resize(), VipsInterpolate.

Parameters

in

input image

 

out

output image

 

a

transformation matrix coefficient

 

b

transformation matrix coefficient

 

c

transformation matrix coefficient

 

d

transformation matrix coefficient

 

...

NULL-terminated list of optional named arguments

 

Returns

0 on success, -1 on error


vips_resize ()

int
vips_resize (VipsImage *in,
             VipsImage **out,
             double scale,
             ...);

Optional arguments:

  • vscale : gdouble vertical scale factor

  • kernel : VipsKernel to reduce with

  • centre : gboolean use centre rather than corner sampling convention

Resize an image.

When downsizing, the image is block-shrunk with vips_shrink(), then the image is shrunk again to the target size with vips_reduce(). How much is done by vips_shrink() vs. vips_reduce() varies with the kernel setting.

vips_resize() normally uses VIPS_KERNEL_LANCZOS3 for the final reduce, you can change this with kernel .

Set centre to use centre rather than corner sampling convention. Centre convention can be useful to match the behaviour of other systems.

When upsizing (scale > 1), the operation uses vips_affine() with a VipsInterpolate selected depending on kernel . It will use VipsInterpolateBicubic for VIPS_KERNEL_CUBIC and above.

vips_resize() normally maintains the image apect ratio. If you set vscale , that factor is used for the vertical scale and scale for the horizontal.

This operation does not change xres or yres. The image resolution needs to be updated by the application.

See also: vips_shrink(), vips_reduce().

Parameters

in

input image

 

out

output image

 

scale

scale factor

 

...

NULL-terminated list of optional named arguments

 

Returns

0 on success, -1 on error


vips_mapim ()

int
vips_mapim (VipsImage *in,
            VipsImage **out,
            VipsImage *index,
            ...);

Optional arguments:

  • interpolate : interpolate pixels with this

This operator resamples in using index to look up pixels. out is the same size as index , with each pixel being fetched from that position in in . That is:

1
out[x, y] = in[index[x, y]]

If index has one band, that band must be complex. Otherwise, index must have two bands of any format. Coordinates in index are in pixels, with (0, 0) being the top-left corner of in , and with y increasing down the image. Use vips_xyz() to build index images.

interpolate defaults to bilinear.

This operation does not change xres or yres. The image resolution needs to be updated by the application.

See vips_maplut() for a 1D equivalent of this operation.

See also: vips_xyz(), vips_affine(), vips_resize(), vips_maplut(), VipsInterpolate.

Parameters

in

input image

 

out

output image

 

index

index image

 

...

NULL-terminated list of optional named arguments

 

Returns

0 on success, -1 on error


vips_quadratic ()

int
vips_quadratic (VipsImage *in,
                VipsImage **out,
                VipsImage *coeff,
                ...);

Optional arguments:

  • interpolate : use this interpolator (default bilinear)

This operation is unfinished and unusable, sorry.

See also: vips_affine().

Parameters

in

input image

 

out

output image

 

coeff

horizontal quadratic

 

...

NULL-terminated list of optional named arguments

 

Returns

0 on success, -1 on error

Types and Values

enum VipsKernel

The resampling kernels vips supports. See vips_reduce(), for example.

The Lanczos kernels vary in size with the downsampling ratio.

Members

VIPS_KERNEL_NEAREST

The nearest pixel to the point.

 

VIPS_KERNEL_LINEAR

Calculate a pixel value using linear interpolation.

 

VIPS_KERNEL_CUBIC

Calculate using a 4-element cubic kernel.

 

VIPS_KERNEL_LANCZOS2

Calculate with a two-lobe Lanczos kernel.

 

VIPS_KERNEL_LANCZOS3

Calculate with a three-lobe Lanczos kernel.

 

VIPS_KERNEL_LAST

   

enum VipsSize

Controls whether an operation should upsize, downsize, or both up and downsize.

See also: vips_thumbnail().

Members

VIPS_SIZE_BOTH

size both up and down

 

VIPS_SIZE_UP

only upsize

 

VIPS_SIZE_DOWN

only downsize

 

VIPS_SIZE_LAST