++ed by:

37 PAUSE users
20 non-PAUSE users.

Arnar Mar Hrafnkelsson


Imager::ImageTypes - Internal image representation information


  use Imager;

  $img = Imager->new(); #  Empty image (size is 0 by 0)
  $img->open(file=>'lena.png',type=>'png'); # Read image from file

  $img = Imager->new(xsize=>400, ysize=>300); # RGB data

  $img = Imager->new(xsize=>400, ysize=>300,  # Grayscale
                     channels=>1);            #

  $img = Imager->new(xsize=>400, ysize=>300,  # RGB with alpha
                     channels=>4);            #
  $img = Imager->new(xsize=>200, ysize=>200,  
                     type=>'paletted');       # paletted image
  $img = Imager->new(xsize=>200, ysize=>200,  
                     bits=>16);               # 16 bits/channel rgb
  $img = Imager->new(xsize=>200, ysize=>200,  
                     bits=>'double');         # 'double' floating point
                                              #  per channel

  $img->img_set(xsize=>500, ysize=>500,       # reset the image object

  # Example getting information about an Imager object

  print "Image information:\n";
  print "Width:        ", $img->getwidth(),    "\n";
  print "Height:       ", $img->getheight(),   "\n";
  print "Channels:     ", $img->getchannels(), "\n";
  print "Bits/Channel: ", $img->bits(),        "\n";
  print "Virtual:      ", $img->virtual() ? "Yes" : "No", "\n";
  my $colorcount = $img->getcolorcount(maxcolors=>512);
        print "Actual number of colors in image: ";
  print defined($colorcount) ? $colorcount : ">512", "\n";
  print "Type:         ", $img->type(),        "\n";

  if ($img->type() eq 'direct') {
    print "Modifiable Channels: ";
    print join " ", map {
      ($img->getmask() & 1<<$_) ? $_ : ()
    } 0..$img->getchannels();
    print "\n";
  } else {
    @colors = $img->getcolors();
    print "Palette size: ".@colors."\n";
    my $mx = @colors > 4 ? 4 : 0+@colors;
    print "First $mx entries:\n";
    for (@colors[0..$mx-1]) {
      my @res = $_->rgba();
      print "(", join(", ", @res[0..$img->getchannels()-1]), ")\n";
  my @tags = $img->tags();
  if (@tags) {
    print "Tags:\n";
    for(@tags) {
      print shift @$_, ": ", join " ", @$_, "\n";
  } else {
    print "No tags in image\n";


Imager supports various internal image representations of images. The two major classes are direct mode and paletted mode. In paletted mode an image has a numbered list of colors and the color of each pixel is determined by an index into the table. In direct mode there is no color palette and each pixel has a seperate value for red green and blue for RGB images. To complicate matters it's possible to have other color spaces than RGB, for example, gray, gray and alpha, or red, green, blue and alpha.

In addition it's possible to have direct type images with 8 bits/channel 16 bits/channel or double/channel (64 bits on many systems).

To query an existing image about it's parameters see the bits(), type(), getwidth(), getheight(), getchannels() and virtual() methods.

The coordinate system in Imager has the origin in the upper left corner, see Imager::Draw for details.

Creating Imager Objects

  $img = Imager->new();
  $img->read(file=>"alligator.ppm") or die $img->errstr;

Here new() creates an empty image with width and height of zero. It's only useful for creating an Imager object to call the read() method on later.

  %opts = (xsize=>300, ysize=>200);
  $img = Imager->new(%opts); # create direct mode RGBA image
  $img = Imager->new(%opts, channels=>4); # create direct mode RGBA image

To create paletted images, set the 'type' parameter to 'paletted':

  $img = Imager->new(xsize=>200, ysize=>200, type=>'paletted');

which creates an image with a maxiumum of 256 colors, which you can change by supplying the maxcolors parameter.

For improved color precision you can use the bits parameter to specify 16 bit per channel:

  $img = Imager->new(xsize=>200, ysize=>200,
                     channels=>3, bits=>16);

or for even more precision:

  $img = Imager->new(xsize=>200, ysize=>200,
                     channels=>3, bits=>'double');

to get an image that uses a double for each channel.

Note that as of this writing all functions should work on images with more than 8-bits/channel, but many will only work at only 8-bit/channel precision.

Currently only 8-bit, 16-bit, and double per channel image types are available, this may change later.


img_set destroys the image data in the object and creates a new one with the given dimensions and channels. For a way to convert image data between formats see the convert() method.

  $img->img_set(xsize=>500, ysize=>500, channels=>4);

Getting Information About an Imager Object

  print "Image width: ", $img->getwidth(), "\n";

The getwidth() method returns the width of the image. This value comes either from new() with xsize,ysize parameters or from reading data from a file with read(). If called on an image that has no valid data in it like Imager->new() returns, the return value of getwidth() is undef.

  print "Image height: ", $img->getheight(), "\n";

Same details apply as for getwidth.

  print "Image has ",$img->getchannels(), " channels\n";

To get the number of channels in an image getchannels() is used.


It is possible to have Imager find the number of colors in an image by with the getcolorcount() method. It requires memory proportionally to the number of colors in the image so it is possible to have it stop sooner if you only need to know if there are more than a certain number of colors in the image. If there are more colors than asked for the function return undef. Examples:

  if (defined($img->getcolorcount(maxcolors=>512)) {
    print "Less than 512 colors in image\n";

The bits() method retrieves the number of bits used to represent each channel in a pixel, 8 for a normal image, 16 for 16-bit image and 'double' for a double/channel image.


The type() method returns either 'direct' for truecolor images or 'paletted' for paletted images.


The virtual() method returns non-zero if the image contains no actual pixels, for example masked images.

Direct Type Images

  @rgbanames = qw( red green blue alpha );
  my $mask = $img->getmask();
  print "Modifiable channels:\n";
  for (0..$img->getchannels()-1) {
    print $rgbanames[$_],"\n" if $mask & 1<<$_;

getmask() is used to fetch the current channel mask. The mask determines what channels are currently modifiable in the image. The channel mask is an integer value, if the i-th lsb is set the i-th channel is modifiable.

  $mask = $img->getmask();
  $img->setmask(mask=>8);     # modify alpha only


  $img->setmask(mask=>$mask); # restore previous mask

setmask() is used to set the channel mask of the image. See getmask for details.

Palette Type Images

In general you can work with paletted images in the same way as RGB images, except that if you attempt to draw to a paletted image with a color that is not in the image's palette, the image will be converted to an RGB image. This means that drawing on a paletted image with anti-aliasing enabled will almost certainly convert the image to RGB.

Palette management takes place through addcolors(), setcolors(), getcolors() and findcolor():


You can add colors to a paletted image with the addcolors() method:

   my @colors = ( Imager::Color->new(255, 0, 0), 
                  Imager::Color->new(0, 255, 0) );
   my $index = $img->addcolors(colors=>\@colors);

The return value is the index of the first color added, or undef if adding the colors would overflow the palette.

  $img->setcolors(start=>$start, colors=>\@colors);

Once you have colors in the palette you can overwrite them with the setcolors() method: sercolors() returns true on success.


To retrieve existing colors from the palette use the getcolors() method:

  # get the whole palette
  my @colors = $img->getcolors();
  # get a single color
  my $color = $img->getcolors(start=>$index);
  # get a range of colors
  my @colors = $img->getcolors(start=>$index, count=>$count);

To quickly find a color in the palette use findcolor():

  my $index = $img->findcolor(color=>$color);

which returns undef on failure, or the index of the color.

You can get the current palette size with $img->colorcount, and the maximum size of the palette with $img->maxcolors.

Conversion Between Image Types

Warning: if you draw on a paletted image with colors that aren't in the palette, the image will be internally converted to a normal image.


You can create a new paletted image from an existing image using the to_paletted() method:

 $palimg = $img->to_paletted(\%opts)

where %opts contains the options specified under "Quantization options".


You can convert a paletted image (or any image) to an 8-bit/channel RGB image with:

  $rgbimg = $img->to_rgb8;

Masked Images

Masked images let you control which pixels are modified in an underlying image. Where the first channel is completely black in the mask image, writes to the underlying image are ignored.

For example, given a base image called $img:

  my $mask = Imager->new(xsize=>$img->getwidth, ysize=>getheight,
  # ... draw something on the mask
  my $maskedimg = $img->masked(mask=>$mask);

You can specifiy the region of the underlying image that is masked using the left, top, right and bottom options.

If you just want a subset of the image, without masking, just specify the region without specifying a mask.


Image tags contain meta-data about the image, ie. information not stored as pixels of the image.

At the perl level each tag has a name or code and a value, which is an integer or an arbitrary string. An image can contain more than one tag with the same name or code.

You can retrieve tags from an image using the tags() method, you can get all of the tags in an image, as a list of array references, with the code or name of the tag followed by the value of the tag:

  my @alltags = $img->tags;

or you can get all tags that have a given name:

  my @namedtags = $img->tags(name=>$name);

or a given code:

  my @tags = $img->tags(code=>$code);

You can add tags using the addtag() method, either by name:

  my $index = $img->addtag(name=>$name, value=>$value);

or by code:

  my $index = $img->addtag(code=>$code, value=>$value);

You can remove tags with the deltag() method, either by index:


or by name:


or by code:


In each case deltag() returns the number of tags deleted.

Common Tags

Many tags are only meaningful for one format. GIF looping information is pretty useless for JPEG for example. Thus, many tags are set by only a single reader or used by a single writer. For a complete list of format specific tags see Imager::Files.

Since tags are a relatively new addition their use is not wide spread but eventually we hope to have all the readers for various formats set some standard information.


The spatial resolution of the image in pixels per inch. If the image format uses a different scale, eg. pixels per meter, then this value is converted. A floating point number stored as a string.


If this is non-zero then the values in i_xres and i_yres are treated as a ratio only. If the image format does not support aspect ratios then this is scaled so the smaller value is 72dpi.


If this tag is present then the whole image could not be read. This isn't implemented for all images yet.

Quantization options

These options can be specified when calling write_multi() for gif files, when writing a single image with the gifquant option set to 'gen', or for direct calls to i_writegif_gen and i_writegif_callback.


A arrayref of colors that are fixed. Note that some color generators will ignore this.


The type of transparency processing to perform for images with an alpha channel where the output format does not have a proper alpha channel (eg. gif). This can be any of:


No transparency processing is done. (default)


Pixels more transparent that tr_threshold are rendered as transparent.


An error diffusion dither is done on the alpha channel. Note that this is independent of the translation performed on the colour channels, so some combinations may cause undesired artifacts.


The ordered dither specified by tr_orddith is performed on the alpha channel.

This will only be used if the image has an alpha channel, and if there is space in the palette for a transparency colour.


The highest alpha value at which a pixel will be made transparent when transp is 'threshold'. (0-255, default 127)


The type of error diffusion to perform on the alpha channel when transp is 'errdiff'. This can be any defined error diffusion type except for custom (see errdiff below).


The type of ordered dither to perform on the alpha channel when transp is 'ordered'. Possible values are:


A semi-random map is used. The map is the same each time.


8x8 dot dither.


4x4 dot dither


horizontal line dither.


vertical line dither.


diagonal line dither


diagonal line dither


dot matrix dither (currently the default). This is probably the best for displays (like web pages).


A custom dither matrix is used - see tr_map


When tr_orddith is custom this defines an 8 x 8 matrix of integers representing the transparency threshold for pixels corresponding to each position. This should be a 64 element array where the first 8 entries correspond to the first row of the matrix. Values should be betweern 0 and 255.


Defines how the quantization engine will build the palette(s). Currently this is ignored if 'translate' is 'giflib', but that may change. Possible values are:


Only colors supplied in 'colors' are used.


The web color map is used (need url here.)


The original code for generating the color map (Addi's code) is used.


Uses a mediancut algorithm, faster than 'addi', but not as good a result.

Other methods may be added in the future.


A arrayref containing Imager::Color objects, which represents the starting set of colors to use in translating the images. webmap will ignore this. The final colors used are copied back into this array (which is expanded if necessary.)


The maximum number of colors to use in the image.


The method used to translate the RGB values in the source image into the colors selected by make_colors. Note that make_colors is ignored whene translate is 'giflib'.

Possible values are:


The giflib native quantization function is used.


The closest color available is used.


The pixel color is modified by perturb, and the closest color is chosen.


An error diffusion dither is performed.

It's possible other transate values will be added.


The type of error diffusion dither to perform. These values (except for custom) can also be used in tr_errdif.


Floyd-Steinberg dither


Jarvis, Judice and Ninke dither


Stucki dither


Custom. If you use this you must also set errdiff_width, errdiff_height and errdiff_map.


When translate is 'errdiff' and errdiff is 'custom' these define a custom error diffusion map. errdiff_width and errdiff_height define the size of the map in the arrayref in errdiff_map. errdiff_orig is an integer which indicates the current pixel position in the top row of the map.


When translate is 'perturb' this is the magnitude of the random bias applied to each channel of the pixel before it is looked up in the color table.

1 POD Error

The following errors were encountered while parsing the POD:

Around line 323:

You forgot a '=back' before '=head2'