Writing a new loader

Add a new loader to VIPS by subclassing VipsForeignLoad. Subclasses need to implement at least header() .

header() must set at least the header fields of out . load() , if defined, must load the pixels to real .

The suffix list is used to select a format to save a file in, and to pick a loader if you don't define is_a().

You should also define nickname and description in VipsObject.

As a complete example, here's code for a PNG loader, minus the actual calls to libpng.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90

typedef struct _VipsForeignLoadPng {
  VipsForeignLoad parent_object;

  char *filename; 
} VipsForeignLoadPng;

typedef VipsForeignLoadClass VipsForeignLoadPngClass;

G_DEFINE_TYPE( VipsForeignLoadPng, vips_foreign_load_png, 
  VIPS_TYPE_FOREIGN_LOAD );

static VipsForeignFlags
vips_foreign_load_png_get_flags_filename( const char *filename )
{
  VipsForeignFlags flags;

  flags = 0;
  if( vips__png_isinterlaced( filename ) )
    flags = VIPS_FOREIGN_PARTIAL;
  else
    flags = VIPS_FOREIGN_SEQUENTIAL;

  return( flags );
}

static VipsForeignFlags
vips_foreign_load_png_get_flags( VipsForeignLoad *load )
{
  VipsForeignLoadPng *png = (VipsForeignLoadPng *) load;

  return( vips_foreign_load_png_get_flags_filename( png->filename ) );
}

static int
vips_foreign_load_png_header( VipsForeignLoad *load )
{
  VipsForeignLoadPng *png = (VipsForeignLoadPng *) load;

  if( vips__png_header( png->filename, load->out ) )
    return( -1 );

  return( 0 );
}

static int
vips_foreign_load_png_load( VipsForeignLoad *load )
{
  VipsForeignLoadPng *png = (VipsForeignLoadPng *) load;

  if( vips__png_read( png->filename, load->real ) )
    return( -1 );

  return( 0 );
}

static void
vips_foreign_load_png_class_init( VipsForeignLoadPngClass *class )
{
  GObjectClass *gobject_class = G_OBJECT_CLASS( class );
  VipsObjectClass *object_class = (VipsObjectClass *) class;
  VipsForeignClass *foreign_class = (VipsForeignClass *) class;
  VipsForeignLoadClass *load_class = (VipsForeignLoadClass *) class;

  gobject_class->set_property = vips_object_set_property;
  gobject_class->get_property = vips_object_get_property;

  object_class->nickname = "pngload";
  object_class->description = _( "load png from file" );

  foreign_class->suffs = vips__png_suffs;

  load_class->is_a = vips__png_ispng;
  load_class->get_flags_filename = 
    vips_foreign_load_png_get_flags_filename;
  load_class->get_flags = vips_foreign_load_png_get_flags;
  load_class->header = vips_foreign_load_png_header;
  load_class->load = vips_foreign_load_png_load;

  VIPS_ARG_STRING( class, "filename", 1, 
    _( "Filename" ),
    _( "Filename to load from" ),
    VIPS_ARGUMENT_REQUIRED_INPUT, 
    G_STRUCT_OFFSET( VipsForeignLoadPng, filename ),
    NULL );
}

static void
vips_foreign_load_png_init( VipsForeignLoadPng *png )
{
}

Writing a new saver

Call your saver in the class' build() method after chaining up. The prepared image should be ready for you to save in ready .

As a complete example, here's the code for the CSV saver, minus the calls to the actual save routines.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69

typedef struct _VipsForeignSaveCsv {
  VipsForeignSave parent_object;

  char *filename; 
  const char *separator;
} VipsForeignSaveCsv;

typedef VipsForeignSaveClass VipsForeignSaveCsvClass;

G_DEFINE_TYPE( VipsForeignSaveCsv, vips_foreign_save_csv, 
  VIPS_TYPE_FOREIGN_SAVE );

static int
vips_foreign_save_csv_build( VipsObject *object )
{
  VipsForeignSave *save = (VipsForeignSave *) object;
  VipsForeignSaveCsv *csv = (VipsForeignSaveCsv *) object;

  if( VIPS_OBJECT_CLASS( vips_foreign_save_csv_parent_class )->
    build( object ) )
    return( -1 );

  if( vips__csv_write( save->ready, csv->filename, csv->separator ) )
    return( -1 );

  return( 0 );
}

static void
vips_foreign_save_csv_class_init( VipsForeignSaveCsvClass *class )
{
  GObjectClass *gobject_class = G_OBJECT_CLASS( class );
  VipsObjectClass *object_class = (VipsObjectClass *) class;
  VipsForeignClass *foreign_class = (VipsForeignClass *) class;
  VipsForeignSaveClass *save_class = (VipsForeignSaveClass *) class;

  gobject_class->set_property = vips_object_set_property;
  gobject_class->get_property = vips_object_get_property;

  object_class->nickname = "csvsave";
  object_class->description = _( "save image to csv file" );
  object_class->build = vips_foreign_save_csv_build;

  foreign_class->suffs = vips__foreign_csv_suffs;

  save_class->saveable = VIPS_SAVEABLE_MONO;
  // no need to define ->format_table, we don't want the input 
  // cast for us

  VIPS_ARG_STRING( class, "filename", 1, 
    _( "Filename" ),
    _( "Filename to save to" ),
    VIPS_ARGUMENT_REQUIRED_INPUT, 
    G_STRUCT_OFFSET( VipsForeignSaveCsv, filename ),
    NULL );

  VIPS_ARG_STRING( class, "separator", 13, 
    _( "Separator" ), 
    _( "Separator characters" ),
    VIPS_ARGUMENT_OPTIONAL_INPUT,
    G_STRUCT_OFFSET( VipsForeignSaveCsv, separator ),
    "\t" ); 
}

static void
vips_foreign_save_csv_init( VipsForeignSaveCsv *csv )
{
  csv->separator = g_strdup( "\t" );
}

vips_foreign_map ()

void *
vips_foreign_map (const char *base,
                  VipsSListMap2Fn fn,
                  void *a,
                  void *b);

Apply a function to every VipsForeignClass that VIPS knows about. Foreigns are presented to the function in priority order.

Like all VIPS map functions, if fn returns NULL, iteration continues. If it returns non-NULL, iteration terminates and that value is returned. The map function returns NULL if all calls return NULL.

See also: vips_slist_map().

vips_foreign_find_load ()

const char *
vips_foreign_find_load (const char *filename);

Searches for an operation you could use to load filename . Any trailing options on filename are stripped and ignored.

See also: vips_foreign_find_load_buffer(), vips_image_new_from_file().

vips_foreign_find_load_buffer ()

const char *
vips_foreign_find_load_buffer (const void *data,
                               size_t size);

Searches for an operation you could use to load a memory buffer. To see the range of buffer loaders supported by your vips, try something like:

vips -l | grep load_buffer

See also: vips_image_new_from_buffer().

vips_foreign_flags ()

VipsForeignFlags
vips_foreign_flags (const char *loader,
                    const char *filename);

Return the flags for filename using loader . loader is something like "tiffload" or "VipsForeignLoadTiff".

vips_foreign_is_a ()

gboolean
vips_foreign_is_a (const char *loader,
                   const char *filename);

Return TRUE if filename can be loaded by loader . loader is something like "tiffload" or "VipsForeignLoadTiff".

vips_foreign_is_a_buffer ()

gboolean
vips_foreign_is_a_buffer (const char *loader,
                          const void *data,
                          size_t size);

Return TRUE if data can be loaded by loader . loader is something like "tiffload_buffer" or "VipsForeignLoadTiffBuffer".

vips_foreign_find_save ()

const char *
vips_foreign_find_save (const char *filename);

Searches for an operation you could use to write to filename . Any trailing options on filename are stripped and ignored.

See also: vips_foreign_find_save_buffer(), vips_image_write_to_file().

vips_foreign_find_save_buffer ()

const char *
vips_foreign_find_save_buffer (const char *suffix);

Searches for an operation you could use to write to a buffer in suffix format.

See also: vips_image_write_to_buffer().

vips_vipsload ()

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

Read in a vips image.

See also: vips_vipssave().

vips_vipssave ()

int
vips_vipssave (VipsImage *in,
               const char *filename,
               ...);

Write in to filename in VIPS format.

See also: vips_vipsload().

vips_openslideload ()

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

Optional arguments:

Read a virtual slide supported by the OpenSlide library into a VIPS image. OpenSlide supports images in Aperio, Hamamatsu, MIRAX, Sakura, Trestle, and Ventana formats.

To facilitate zooming, virtual slide formats include multiple scaled-down versions of the high-resolution image. These are typically called "levels". By default, vips_openslideload() reads the highest-resolution level (level 0). Set level to the level number you want.

In addition to the slide image itself, virtual slide formats sometimes include additional images, such as a scan of the slide's barcode. OpenSlide calls these "associated images". To read an associated image, set associated to the image's name. A slide's associated images are listed in the "slide-associated-images" metadata item.

The output of this operator is always RGBA.

See also: vips_image_new_from_file().

vips_jpegload ()

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

Optional arguments:

Read a JPEG file into a VIPS image. It can read most 8-bit JPEG images, including CMYK and YCbCr.

shrink means shrink by this integer factor during load. Possible values are 1, 2, 4 and 8. Shrinking during read is very much faster than decompressing the whole image and then shrinking later.

Setting fail to TRUE makes the JPEG reader fail on any warnings. This can be useful for detecting truncated files, for example. Normally reading these produces a warning, but no fatal error.

Setting autorotate to TRUE will make the loader interpret the orientation tag and automatically rotate the image appropriately during load.

If autorotate is FALSE, the metadata field VIPS_META_ORIENTATION is set to the value of the orientation tag. Applications may read and interpret this field as they wish later in processing. See vips_autorot(). Save operations will use VIPS_META_ORIENTATION, if present, to set the orientation of output images.

Example:

1
2
3
4

vips_jpegload( "fred.jpg", &out,
    "shrink", 8,
    "fail", TRUE,
    NULL );

Any embedded ICC profiles are ignored: you always just get the RGB from the file. Instead, the embedded profile will be attached to the image as VIPS_META_ICC_NAME. You need to use something like vips_icc_import() to get CIE values from the file.

EXIF metadata is attached as VIPS_META_EXIF_NAME, IPCT as VIPS_META_IPCT_NAME, and XMP as VIPS_META_XMP_NAME.

The int metadata item "jpeg-multiscan" is set to the result of jpeg_has_multiple_scans(). Interlaced jpeg images need a large amount of memory to load, so this field gives callers a chance to handle these images differently.

The EXIF thumbnail, if present, is attached to the image as "jpeg-thumbnail-data". See vips_image_get_blob().

See also: vips_jpegload_buffer(), vips_image_new_from_file(), vips_autorot().

vips_jpegload_buffer ()

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

Optional arguments:

Read a JPEG-formatted memory block into a VIPS image. Exactly as vips_jpegload(), but read from a memory buffer.

You must not free the buffer while out is active. The “postclose” signal on out is a good place to free.

See also: vips_jpegload().

vips_jpegsave ()

int
vips_jpegsave (VipsImage *in,
               const char *filename,
               ...);

Optional arguments:

Write a VIPS image to a file as JPEG.

Use Q to set the JPEG compression factor. Default 75.

Use profile to give the filename of a profile to be embedded in the JPEG. This does not affect the pixels which are written, just the way they are tagged. You can use the special string "none" to mean "don't attach a profile".

If no profile is specified and the VIPS header contains an ICC profile named VIPS_META_ICC_NAME, the profile from the VIPS header will be attached.

If optimize_coding is set, the Huffman tables are optimised. This is sllightly slower and produces slightly smaller files.

If interlace is set, the jpeg files will be interlaced (progressive jpeg, in jpg parlance). These files may be better for display over a slow network conection, but need much more memory to encode and decode.

If strip is set, no EXIF data, IPCT data, ICC profile or XMP metadata is written into the output file.

If no_subsample is set, chrominance subsampling is disabled. This will improve quality at the cost of larger file size. Useful for high Q factors.

If trellis_quant is set and the version of libjpeg supports it (e.g. mozjpeg >= 3.0), apply trellis quantisation to each 8x8 block. Reduces file size but increases compression time.

If overshoot_deringing is set and the version of libjpeg supports it (e.g. mozjpeg >= 3.0), apply overshooting to samples with extreme values for example 0 and 255 for 8-bit. Overshooting may reduce ringing artifacts from compression, in particular in areas where black text appears on a white background.

If optimize_scans is set and the version of libjpeg supports it (e.g. mozjpeg >= 3.0), split the spectrum of DCT coefficients into separate scans. Reduces file size but increases compression time.

If quant_table is set and the version of libjpeg supports it (e.g. mozjpeg >= 3.0) it selects the quantization table to use:

Quantization table 0 is the default in vips and libjpeg(-turbo), but it tends to favor detail over color accuracy, producting colored patches and stripes as well as heavy banding in flat areas at high compression ratios. Quantization table 2 is a good candidate to try if the default quantization table produces banding or color shifts and is well suited for hires images. Quantization table 3 is the default in mozjpeg and has been tuned to produce good results at the default quality setting; banding at high compression. Quantization table 4 is the most accurate at the cost of compression ratio. Tables 5-7 are based on older research papers, but generally achieve worse compression ratios and/or quality than 2 or 4.

The image is automatically converted to RGB, Monochrome or CMYK before saving.

EXIF data is constructed from VIPS_META_EXIF_NAME, then modified with any other related tags on the image before being written to the file. VIPS_META_RESOLUTION_UNIT is used to set the EXIF resolution unit. VIPS_META_ORIENTATION is used to set the EXIF orientation tag.

IPCT as VIPS_META_IPCT_NAME and XMP as VIPS_META_XMP_NAME are coded and attached.

See also: vips_jpegsave_buffer(), vips_image_write_to_file().

vips_jpegsave_buffer ()

int
vips_jpegsave_buffer (VipsImage *in,
                      void **buf,
                      size_t *len,
                      ...);

Optional arguments:

As vips_jpegsave(), but save to a memory buffer.

The address of the buffer is returned in obuf , the length of the buffer in olen . You are responsible for freeing the buffer with g_free() when you are done with it.

See also: vips_jpegsave(), vips_image_write_to_file().

vips_jpegsave_mime ()

int
vips_jpegsave_mime (VipsImage *in,
                    ...);

Optional arguments:

As vips_jpegsave(), but save as a mime jpeg on stdout.

See also: vips_jpegsave(), vips_image_write_to_file().

vips_webpload ()

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

Optional arguments:

Read a WebP file into a VIPS image.

Use shrink to specify a shrink-on-load factor.

See also: vips_image_new_from_file().

vips_webpload_buffer ()

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

Optional arguments:

Read a WebP-formatted memory block into a VIPS image. Exactly as vips_webpload(), but read from a memory buffer.

You must not free the buffer while out is active. The “postclose” signal on out is a good place to free.

See also: vips_webpload()

vips_webpsave ()

int
vips_webpsave (VipsImage *in,
               const char *filename,
               ...);

Optional arguments:

Write an image to a file in WebP format.

By default, images are saved in lossy format, with Q giving the WebP quality factor. It has the range 0 - 100, with the default 75.

Use preset to hint the image type to the lossy compressor. The default is VIPS_FOREIGN_WEBP_PRESET_DEFAULT. Set smart_subsample to enable high quality chroma subsampling. Use alpha_q to set the quality for the alpha channel in lossy mode. It has the range 1 - 100, with the default 100.

Set lossless to use lossless compression, or combine near_lossless with Q 80, 60, 40 or 20 to apply increasing amounts of preprocessing which improves the near-lossless compression ratio by up to 50%.

See also: vips_webpload(), vips_image_write_to_file().

vips_webpsave_buffer ()

int
vips_webpsave_buffer (VipsImage *in,
                      void **buf,
                      size_t *len,
                      ...);

Optional arguments:

As vips_webpsave(), but save to a memory buffer.

The address of the buffer is returned in buf , the length of the buffer in len . You are responsible for freeing the buffer with g_free() when you are done with it.

See also: vips_webpsave().

vips_webpsave_mime ()

int
vips_webpsave_mime (VipsImage *in,
                    ...);

Optional arguments:

As vips_webpsave(), but save as a mime webp on stdout.

See also: vips_webpsave(), vips_image_write_to_file().

vips_tiffload ()

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

Optional arguments:

Read a TIFF file into a VIPS image. It is a full baseline TIFF 6 reader, with extensions for tiled images, multipage images, LAB colour space, pyramidal images and JPEG compression. including CMYK and YCbCr.

page means load this page from the file. By default the first page (page 0) is read.

Setting autorotate to TRUE will make the loader interpret the orientation tag and automatically rotate the image appropriately during load.

If autorotate is FALSE, the metadata field VIPS_META_ORIENTATION is set to the value of the orientation tag. Applications may read and interpret this field as they wish later in processing. See vips_autorot(). Save operations will use VIPS_META_ORIENTATION, if present, to set the orientation of output images.

Any ICC profile is read and attached to the VIPS image as VIPS_META_ICC_NAME. Any XMP metadata is read and attached to the image as VIPS_META_XMP_NAME. Any IPCT is attached as VIPS_META_IPCT_NAME. The image description is attached as VIPS_META_IMAGEDESCRIPTION. Data in the photoshop tag is attached as VIPS_META_PHOTOSHOP_NAME.

See also: vips_image_new_from_file(), vips_autorot().

vips_tiffload_buffer ()

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

Optional arguments:

Read a TIFF-formatted memory block into a VIPS image. Exactly as vips_tiffload(), but read from a memory source.

You must not free the buffer while out is active. The “postclose” signal on out is a good place to free.

See also: vips_tiffload().

vips_tiffsave ()

int
vips_tiffsave (VipsImage *in,
               const char *filename,
               ...);

Optional arguments:

Write a VIPS image to a file as TIFF.

Use compression to set the tiff compression. Currently jpeg, packbits, fax4, lzw, none and deflate are supported. The default is no compression. JPEG compression is a good lossy compressor for photographs, packbits is good for 1-bit images, and deflate is the best lossless compression TIFF can do.

Use Q to set the JPEG compression factor. Default 75.

Use predictor to set the predictor for lzw and deflate compression.

Predictor is not set by default. There are three predictor values recognised at the moment (2007, July): 1 is no prediction, 2 is a horizontal differencing and 3 is a floating point predictor. Refer to the libtiff specifications for further discussion of various predictors. In short, predictor helps to better compress image, especially in case of digital photos or scanned images and bit depths > 8. Try it to find whether it works for your images.

Use profile to give the filename of a profile to be embedded in the TIFF. This does not affect the pixels which are written, just the way they are tagged. You can use the special string "none" to mean "don't attach a profile".

If no profile is specified and the VIPS header contains an ICC profile named VIPS_META_ICC_NAME, the profile from the VIPS header will be attached.

Set tile to TRUE to write a tiled tiff. By default tiff are written in strips. Use tile_width and tile_height to set the tile size. The defaiult is 128 by 128.

Set pyramid to write the image as a set of images, one per page, of decreasing size.

Set squash to make 8-bit uchar images write as 1-bit TIFFs. Values >128 are written as white, values <=128 as black. Normally vips will write MINISBLACK TIFFs where black is a 0 bit, but if you set miniswhite , it will use 0 for a white bit. Many pre-press applications only work with images which use this sense. miniswhite only affects one-bit images, it does nothing for greyscale images.

Use resunit to override the default resolution unit. The default resolution unit is taken from the header field VIPS_META_RESOLUTION_UNIT. If this field is not set, then VIPS defaults to cm.

Use xres and yres to override the default horizontal and vertical resolutions. By default these values are taken from the VIPS image header. libvips resolution is always in pixels per millimetre.

Set bigtiff to attempt to write a bigtiff. Bigtiff is a variant of the TIFF format that allows more than 4GB in a file.

Set properties to write all vips metadata to the IMAGEDESCRIPTION tag as xml. If properties is not set, the value of VIPS_META_IMAGEDESCRIPTION is used instead.

The value of VIPS_META_XMP_NAME is written to the XMP tag. VIPS_META_ORIENTATION (if set) is used to set the value of the orientation tag. VIPS_META_IPCT (if set) is used to set the value of the IPCT tag. VIPS_META_PHOTOSHOP_NAME (if set) is used to set the value of the PHOTOSHOP tag.

See also: vips_tiffload(), vips_image_write_to_file().

vips_openexrload ()

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

Read a OpenEXR file into a VIPS image.

The reader can handle scanline and tiled OpenEXR images. It can't handle OpenEXR colour management, image attributes, many pixel formats, anything other than RGBA.

This reader uses the rather limited OpenEXR C API. It should really be redone in C++.

See also: vips_image_new_from_file().

vips_fitsload ()

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

Read a FITS image file into a VIPS image.

This operation can read images with up to three dimensions. Any higher dimensions must be empty.

It can read 8, 16 and 32-bit integer images, signed and unsigned, float and double.

FITS metadata is attached with the "fits-" prefix.

See also: vips_image_new_from_file().

vips_fitssave ()

int
vips_fitssave (VipsImage *in,
               const char *filename,
               ...);

Write a VIPS image to a file in FITS format.

See also: vips_image_write_to_file().

vips_analyzeload ()

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

Load an Analyze 6.0 file. If filename is "fred.img", this will look for an image header called "fred.hdr" and pixel data in "fred.img". You can also load "fred" or "fred.hdr".

Images are loaded lazilly and byte-swapped, if necessary. The Analyze metadata is read and attached.

See also: vips_image_new_from_file().

vips_rawload ()

int
vips_rawload (const char *filename,
              VipsImage **out,
              int width,
              int height,
              int bands,
              ...);

Optional arguments:

This operation mmaps the file, setting up out so that access to that image will read from the file.

out will be a 8-bit uchar image with bands image bands, so bands can be thought of as meaning "number of bytes per pixel". Use functions like vips_copy() to set the exact band format, number of bands, and so on. Use vips_byteswap() to reverse the byte ordering.

See also: vips_image_new_from_file(), vips_copy(), vips_byteswap().

vips_rawsave ()

int
vips_rawsave (VipsImage *in,
              const char *filename,
              ...);

Writes the pixels in in to the file filename with no header or other metadata.

See also: vips_image_write_to_file().

vips_rawsave_fd ()

int
vips_rawsave_fd (VipsImage *in,
                 int fd,
                 ...);

Writes the pixels in in to the fd with no header or other metadata. Handy for implementing other savers.

See also: vips_rawsave().

vips_csvload ()

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

Optional arguments:

Load a CSV (comma-separated values) file. The output image is always 1 band (monochrome), VIPS_FORMAT_DOUBLE. Use vips_bandfold() to turn RGBRGBRGB mono images into colour iamges.

Items in lines can be either floating point numbers in the C locale, or strings enclosed in double-quotes ("), or empty. You can use a backslash() within the quotes to escape special characters, such as quote marks.

The reader is deliberately rather fussy: it will fail if there are any short lines, or if the file is too short. It will ignore lines that are too long.

skip sets the number of lines to skip at the start of the file. Default zero.

lines sets the number of lines to read from the file. Default -1, meaning read all lines to end of file.

whitespace sets the skippable whitespace characters. Default space. Whitespace characters are always run together.

separator sets the characters that separate fields. Default ;,tab. Separators are never run together.

See also: vips_image_new_from_file(), vips_bandfold().

vips_csvsave ()

int
vips_csvsave (VipsImage *in,
              const char *filename,
              ...);

Optional arguments:

Writes the pixels in in to the filename as CSV (comma-separated values). The image is written one line of text per scanline. Complex numbers are written as "(real,imaginary)" and will need extra parsing I guess. Only the first band is written.

separator gives the string to use to separate numbers in the output. The default is "\t" (tab).

See also: vips_image_write_to_file().

vips_matrixload ()

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

Reads a matrix from a file.

Matrix files have a simple format that's supposed to be easy to create with a text editor or a spreadsheet.

The first line has four numbers for width, height, scale and offset (scale and offset may be omitted, in which case they default to 1.0 and 0.0). Scale must be non-zero. Width and height must be positive integers. The numbers are separated by any mixture of spaces, commas, tabs and quotation marks ("). The scale and offset fields may be floating-point, and must use '.' as a decimal separator.

Subsequent lines each hold one row of matrix data, with numbers again separated by any mixture of spaces, commas, tabs and quotation marks ("). The numbers may be floating-point, and must use '.' as a decimal separator.

Extra characters at the ends of lines or at the end of the file are ignored.

See also: vips_csvload().

vips_matrixsave ()

int
vips_matrixsave (VipsImage *in,
                 const char *filename,
                 ...);

Write in to filename in matrix format. See vips_matrixload() for a description of the format.

See also: vips_matrixload().

vips_matrixprint ()

int
vips_matrixprint (VipsImage *in,
                  ...);

Print in to stdout in matrix format. See vips_matrixload() for a description of the format.

See also: vips_matrixload().

vips_magickload ()

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

Optional arguments:

Read in an image using libMagick, the ImageMagick library. This library can read more than 80 file formats, including SVG, BMP, EPS, DICOM and many others. The reader can handle any ImageMagick image, including the float and double formats. It will work with any quantum size, including HDR. Any metadata attached to the libMagick image is copied on to the VIPS image.

The reader should also work with most versions of GraphicsMagick. See the "--with-magickpackage" configure option.

Normally it will only load the first image in a many-image sequence (such as a GIF). Set all_frames to true to read the whole image sequence.

density is "WxH" in DPI, e.g. "600x300" or "600" (default is "72x72"). See the density docs on the imagemagick website.

See also: vips_image_new_from_file().

vips_magickload_buffer ()

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

Optional arguments:

Read an image memory block using libMagick into a VIPS image. Exactly as vips_magickload(), but read from a memory source.

You must not free the buffer while out is active. The “postclose” signal on out is a good place to free.

See also: vips_magickload().

vips_pngload ()

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

Read a PNG file into a VIPS image. It can read all png images, including 8- and 16-bit images, 1 and 3 channel, with and without an alpha channel.

Any ICC profile is read and attached to the VIPS image.

See also: vips_image_new_from_file().

vips_pngload_buffer ()

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

Read a PNG-formatted memory block into a VIPS image. It can read all png images, including 8- and 16-bit images, 1 and 3 channel, with and without an alpha channel.

Any ICC profile is read and attached to the VIPS image.

You must not free the buffer while out is active. The “postclose” signal on out is a good place to free.

See also: vips_pngload().

vips_pngsave ()

int
vips_pngsave (VipsImage *in,
              const char *filename,
              ...);

Optional arguments:

Write a VIPS image to a file as PNG.

compression means compress with this much effort (0 - 9). Default 6.

Set interlace to TRUE to interlace the image with ADAM7 interlacing. Beware than an interlaced PNG can be up to 7 times slower to write than a non-interlaced image.

Use profile to give the filename of a profile to be embedded in the PNG. This does not affect the pixels which are written, just the way they are tagged. You can use the special string "none" to mean "don't attach a profile".

If profile is specified and the VIPS header contains an ICC profile named VIPS_META_ICC_NAME ("icc-profile-data"), the profile from the VIPS header will be attached.

Use filter to specify one or more filters (instead of adaptive filtering), see VipsForeignPngFilter.

The image is automatically converted to RGB, RGBA, Monochrome or Mono + alpha before saving. Images with more than one byte per band element are saved as 16-bit PNG, others are saved as 8-bit PNG.

See also: vips_image_new_from_file().

vips_pngsave_buffer ()

int
vips_pngsave_buffer (VipsImage *in,
                     void **buf,
                     size_t *len,
                     ...);

Optional arguments:

As vips_pngsave(), but save to a memory buffer.

The address of the buffer is returned in buf , the length of the buffer in len . You are responsible for freeing the buffer with g_free() when you are done with it.

See also: vips_pngsave(), vips_image_write_to_file().

vips_ppmload ()

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

Read a PPM/PBM/PGM/PFM file into a VIPS image.

It can read 1, 8, 16 and 32 bit images, colour or monochrome, stored in binary or in ASCII. One bit images become 8 bit VIPS images, with 0 and 255 for 0 and 1.

See also: vips_image_new_from_file().

vips_ppmsave ()

int
vips_ppmsave (VipsImage *in,
              const char *filename,
              ...);

Optional arguments:

Write a VIPS image to a file as PPM. It can write 1, 8, 16 or 32 bit unsigned integer images, float images, colour or monochrome, stored as binary or ASCII. Integer images of more than 8 bits can only be stored in ASCII.

When writing float (PFM) images the scale factor is set from the "pfm-scale" metadata.

Set ascii to TRUE to write as human-readable ASCII. Normally data is written in binary.

Set squash to TRUE to squash 8-bit images down to one bit. The saver does no dithering, that's up to you.

See also: vips_image_write_to_file().

vips_matload ()

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

Read a Matlab save file into a VIPS image.

This operation searches the save file for the first array variable with between 1 and 3 dimensions and loads it as an image. It will not handle complex images. It does not handle sparse matrices.

See also: vips_image_new_from_file().

vips_radload ()

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

Read a Radiance (HDR) file into a VIPS image.

Radiance files are read as VIPS_CODING_RAD. They have one byte for each of red, green and blue, and one byte of shared exponent. Some operations (like vips_extract_area()) can work directly with images in this format, but mmany (all the arithmetic operations, for example) will not. Unpack VIPS_CODING_RAD images to 3 band float with vips_rad2float() if you want to do arithmetic on them.

This operation ignores some header fields, like VIEW and DATE. It will not rotate/flip as the FORMAT string asks.

Sections of this reader from Greg Ward and Radiance with kind permission.

See also: vips_image_new_from_file().

vips_radsave ()

int
vips_radsave (VipsImage *in,
              const char *filename,
              ...);

Write a VIPS image in Radiance (HDR) format.

Sections of this reader from Greg Ward and Radiance with kind permission.

See also: vips_image_write_to_file().

vips_radsave_buffer ()

int
vips_radsave_buffer (VipsImage *in,
                     void **buf,
                     size_t *len,
                     ...);

As vips_radsave(), but save to a memory buffer.

The address of the buffer is returned in buf , the length of the buffer in len . You are responsible for freeing the buffer with g_free() when you are done with it.

See also: vips_radsave(), vips_image_write_to_file().

vips_pdfload ()

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

Optional arguments:

Render a PDF file into a VIPS image. Rendering uses the libpoppler library and should be fast.

The output image is always RGBA --- CMYK PDFs will be converted. If you need CMYK bitmaps, you should use vips_magickload() instead.

Rendering is progressive, that is, the image is rendered in strips equal in height to the tile height. If your PDF contains large image files and they span several strips in the output image, they will be decoded multiple times. To fix this, increase the the tile height, for example:

1

vips copy huge.pdf x.png --vips-tile-height=1024

Will process images in 1024-pixel high strips, potentially much faster, though of course also using a lot more memory.

Use page to select a page to render, numbering from zero.

Use n to select the number of pages to render. The default is 1. Pages are rendered in a vertical column, with each individual page aligned to the left. Set to -1 to mean "until the end of the document". Use vips_grid() to change page layout.

Use dpi to set the rendering resolution. The default is 72. Alternatively, you can scale the rendering from the default 1 point == 1 pixel by setting scale .

The operation fills a number of header fields with metadata, for example "pdf-author". They may be useful.

This function only reads the image header and does not render any pixel data. Rendering occurs when pixels are accessed.

See also: vips_image_new_from_file(), vips_magickload().

vips_pdfload_buffer ()

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

Optional arguments:

Read a PDF-formatted memory block into a VIPS image. Exactly as vips_pdfload(), but read from a memory buffer.

You must not free the buffer while out is active. The “postclose” signal on out is a good place to free.

See also: vips_pdfload().

vips_svgload ()

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

Optional arguments:

Render a SVG file into a VIPS image. Rendering uses the librsvg library and should be fast.

Use dpi to set the rendering resolution. The default is 72. Alternatively, you can scale the rendering from the default 1 point == 1 pixel by scale .

This function only reads the image header and does not render any pixel data. Rendering occurs when pixels are accessed.

See also: vips_image_new_from_file().

vips_svgload_buffer ()

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

Optional arguments:

Read a SVG-formatted memory block into a VIPS image. Exactly as vips_svgload(), but read from a memory buffer.

You must not free the buffer while out is active. The “postclose” signal on out is a good place to free.

See also: vips_svgload().

vips_gifload ()

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

Optional arguments:

Read a GIF file into a VIPS image. Rendering uses the giflib library.

Use page to set page number (frame number) to read.

The whole GIF is rendered into memory on header access. The output image will be 1, 2, 3 or 4 bands depending on what the reader finds in the file.

See also: vips_image_new_from_file().

vips_gifload_buffer ()

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

Optional arguments:

Read a GIF-formatted memory block into a VIPS image. Exactly as vips_gifload(), but read from a memory buffer.

You must not free the buffer while out is active. The “postclose” signal on out is a good place to free.

See also: vips_gifload().

vips_dzsave ()

int
vips_dzsave (VipsImage *in,
             const char *name,
             ...);

Optional arguments:

Save an image as a set of tiles at various resolutions. By default dzsave uses DeepZoom layout -- use layout to pick other conventions.

vips_dzsave() creates a directory called name to hold the tiles. If name ends .zip, vips_dzsave() will create a zip file called name to hold the tiles. You can use container to force zip file output.

You can set suffix to something like ".jpg[Q=85]" to control the tile write options.

In Google layout mode, edge tiles are expanded to tile_size by tile_size pixels. Normally they are filled with white, but you can set another colour with background . Images are usually placed at the top-left of the tile, but you can have them centred by turning on centre .

You can set the size and overlap of tiles with tile_size and overlap . They default to the correct settings for the selected layout . The deepzoom defaults produce 256x256 jpeg files for centre tiles, the most efficient size.

Use depth to control how low the pyramid goes. This defaults to the correct setting for the layout you select.

If properties is TRUE, vips_dzsave() will write a file called vips-properties.xml to the output directory. This file lists all of the metadata attached to in in an obvious manner. It can be useful for viewing programs which wish to use fields from source files loaded via vips_openslideload().

If container is set to zip, you can set a compression level from -1 (use zlib default), 0 (store, compression disabled) to 9 (max compression). If no value is given, the default is to store files without compression.

See also: vips_tiffsave().

enum VipsForeignFlags

Some hints about the image loader.

VIPS_FOREIGN_PARTIAL means that the image can be read directly from the file without needing to be unpacked to a temporary image first.

VIPS_FOREIGN_SEQUENTIAL means that the loader supports lazy reading, but only top-to-bottom (sequential) access. Formats like PNG can read sets of scanlines, for example, but only in order.

If neither PARTIAL or SEQUENTIAL is set, the loader only supports whole image read. Setting both PARTIAL and SEQUENTIAL is an error.

VIPS_FOREIGN_BIGENDIAN means that image pixels are most-significant byte first. Depending on the native byte order of the host machine, you may need to swap bytes. See vips_copy().

enum VipsSaveable

See also: VipsForeignSave.

enum VipsForeignWebpPreset

Tune lossy encoder settings for different image types.

enum VipsForeignTiffCompression

The compression types supported by the tiff writer.

Use Q to set the jpeg compression level, default 75.

Use prediction to set the lzw or deflate prediction, default none.

enum VipsForeignTiffPredictor

The predictor can help deflate and lzw compression. The values are fixed by the tiff library.

enum VipsForeignTiffResunit

Use inches or centimeters as the resolution unit for a tiff file.

enum VipsForeignPngFilter

http://www.w3.org/TR/PNG-Filters.html The values mirror those of png.h in libpng.

enum VipsForeignDzLayout

What directory layout and metadata standard to use.

enum VipsForeignDzDepth

How many pyramid layers to create.

enum VipsForeignDzContainer

How many pyramid layers to create.

The “access” property

  “access”                   VipsAccess

Required access pattern for this file.

Flags: Read / Write

Default value: VIPS_ACCESS_RANDOM

The “disc” property

  “disc”                     gboolean

Open to disc.

Flags: Read / Write

Default value: TRUE

The “flags” property

  “flags”                    VipsForeignFlags

Flags for this file.

Flags: Read / Write

The “out” property

  “out”                      VipsImage *

Output image.

Flags: Read / Write

The “sequential” property

  “sequential”               gboolean

Sequential read only.

Flags: Read / Write

Default value: FALSE

The “background” property

  “background”               VipsArrayDouble *

Background value.

Flags: Read / Write

The “in” property

  “in”                       VipsImage *

Image to save.

Flags: Read / Write

The “strip” property

  “strip”                    gboolean

Strip all metadata from image.

Flags: Read / Write

Default value: FALSE