Imager::Transformations(3) Simple transformations of one image into another.

SYNOPSIS


use Imager;
$newimg = $img->copy();
$newimg = $img->scale(xpixels=>400, qtype => 'mixing');
$newimg = $img->scale(xpixels=>400, ypixels=>400);
$newimg = $img->scale(xpixels=>400, ypixels=>400, type=>'min');
$newimg = $img->scale(scalefactor=>0.25);
$newimg = $img->scaleX(pixels=>400);
$newimg = $img->scaleX(scalefactor=>0.25);
$newimg = $img->scaleY(pixels=>400);
$newimg = $img->scaleY(scalefactor=>0.25);
$newimg = $img->crop(left=>50, right=>100, top=>10, bottom=>100);
$newimg = $img->crop(left=>50, top=>10, width=>50, height=>90);
$dest->paste(left=>40,top=>20,img=>$logo);
$img->rubthrough(src=>$srcimage,tx=>30, ty=>50);
$img->rubthrough(src=>$srcimage,tx=>30, ty=>50,
src_minx=>20, src_miny=>30,
src_maxx=>20, src_maxy=>30);
$img->compose(src => $src, tx => 30, ty => 20, combine => 'color');
$img->compose(src => $src, tx => 30, ty => 20, combine => 'color');
mask => $mask, opacity => 0.5);
$img->flip(dir=>"h"); # horizontal flip
$img->flip(dir=>"vh"); # vertical and horizontal flip
$newimg = $img->copy->flip(dir=>"v"); # make a copy and flip it vertically
my $rot20 = $img->rotate(degrees=>20);
my $rotpi4 = $img->rotate(radians=>3.14159265/4);
# Convert image to gray
$new = $img->convert(preset=>'grey');
# Swap red/green channel
$new = $img->convert(matrix=>[ [ 0, 1, 0 ],
[ 1, 0, 0 ],
[ 0, 0, 1 ] ]);
# build an image using channels from multiple input images
$new = $img->combine(src => [ $im1, $im2, $im3 ]);
$new = $img->combine(src => [ $im1, $im2, $im3 ],
channels => [ 2, 1, 0 ]);
# limit the range of red channel from 0..255 to 0..127
@map = map { int( $_/2 } 0..255;
$img->map( red=>\@map );
# Apply a Gamma of 1.4
my $gamma = 1.4;
my @map = map { int( 0.5 + 255*($_/255)**$gamma ) } 0..255;
$img->map(all=>\@map); # inplace conversion

DESCRIPTION

The methods described in Imager::Transformations fall into two categories. Either they take an existing image and modify it in place, or they return a modified copy.

Functions that modify inplace are "flip()", "paste()", "rubthrough()" and "compose()". If the original is to be left intact it's possible to make a copy and alter the copy:

  $flipped = $img->copy()->flip(dir=>'h');

Image copying/resizing/cropping/rotating

A list of the transformations that do not alter the source image follows:
copy()
To create a copy of an image use the "copy()" method. This is useful if you want to keep an original after doing something that changes the image.

  $newimg = $orig->copy();
scale()
To scale an image so proportions are maintained use the "$img->scale()" method. if you give either a "xpixels" or "ypixels" parameter they will determine the width or height respectively. If both are given the one resulting in a larger image is used, unless you set the "type" parameter to 'min'. example: $img is 700 pixels wide and 500 pixels tall.

  $newimg = $img->scale(xpixels=>400); # 400x285
  $newimg = $img->scale(ypixels=>400); # 560x400
  $newimg = $img->scale(xpixels=>400,ypixels=>400); # 560x400
  $newimg = $img->scale(xpixels=>400,ypixels=>400,type=>'min'); # 400x285
  $newimg = $img->scale(xpixels=>400, ypixels=>400),type=>'nonprop'); # 400x400
  $newimg = $img->scale(scalefactor=>0.25); 175x125 
  $newimg = $img->scale(); # 350x250

If you want to create low quality previews of images you can pass "qtype=>'preview'" to scale and it will use nearest neighbor sampling instead of filtering. It is much faster but also generates worse looking images - especially if the original has a lot of sharp variations and the scaled image is by more than 3-5 times smaller than the original.

"xpixels", "ypixels" - desired size of the scaled image. The "type" parameter controls whether the larger or smaller of the two possible sizes is chosen, or if the image is scaled non-proportionally.
"constrain" - an Image::Math::Constrain object defining the way in which the image size should be constrained.
"scalefactor" - if none of "xpixels", "ypixels", "xscalefactor", "yscalefactor" or "constrain" is supplied then this is used as the ratio to scale by. Default: 0.5.
"xscalefactor", "yscalefactor" - if both are supplied then the image is scaled as per these parameters, whether this is proportionally or not. New in Imager 0.54.
"type" - controls whether the larger or smaller of the two possible sizes is chosen, possible values are:
  • "min" - the smaller of the 2 sizes are chosen.
  • "max" - the larger of the 2 sizes. This is the default.
  • "nonprop" - non-proportional scaling. New in Imager 0.54.

scale() will fail if "type" is set to some other value.

For example, if the original image is 400 pixels wide by 200 pixels high and "xpixels" is set to 300, and "ypixels" is set to 160. When "type" is 'min' the resulting image is 300 x 150, when "type" is 'max' the resulting image is 320 x 160.

"type" is only used if both "xpixels" and "ypixels" are supplied.

"qtype" - defines the quality of scaling performed. Possible values are:
  • "normal" - high quality scaling. This is the default.
  • "preview" - lower quality. When scaling down this will skip input pixels, eg. scaling by 0.5 will skip every other pixel. When scaling up this will duplicate pixels.
  • "mixing" - implements the mixing algorithm implemented by pnmscale. This retains more detail when scaling down than "normal". When scaling down this proportionally accumulates sample data from the pixels, resulting in a proportional mix of all of the pixels. When scaling up this will mix pixels when the sampling grid crosses a pixel boundary but will otherwise copy pixel values.

scale() will fail if "qtype" is set to some other value.

"preview" is faster than "mixing" which is much faster than "normal".

To scale an image on a given axis without maintaining proportions, it is best to call the scaleX() and scaleY() methods with the required dimensions. eg.

  my $scaled = $img->scaleX(pixels=>400)->scaleY(pixels=>200);

From Imager 0.54 you can scale without maintaining proportions either by supplying both the "xscalefactor" and "yscalefactor" arguments:

  my $scaled = $img->scale(xscalefactor => 0.5, yscalefactor => 0.67);

or by supplying "xpixels" and "ypixels" and setting "type" to <nonprop>:

  my $scaled = $im->scale(xpixels => 200, ypixels => 200, type => 'nonprop');

Returns a new scaled image on success. The source image is not modified.

Returns false on failure, check the errstr() method for the reason for failure.

A mandatory warning is produced if scale() is called in void context.

  # setup
  my $image = Imager->new;
  $image->read(file => 'somefile.jpg')
    or die $image->errstr;
  # all full quality unless indicated otherwise
  # half the size:
  my $half = $image->scale;
  # double the size
  my $double = $image->scale(scalefactor => 2.0);
  # so a 400 x 400 box fits in the resulting image:
  my $fit400x400inside = $image->scale(xpixels => 400, ypixels => 400);
  my $fit400x400inside2 = $image->scale(xpixels => 400, ypixels => 400,
                                        type=>'max');
  # fit inside a 400 x 400 box
  my $inside400x400 = $image->scale(xpixels => 400, ypixels => 400,
                              type=>'min');
  # make it 400 pixels wide or high
  my $width400 = $image->scale(xpixels => 400);
  my $height400 = $image->scale(ypixels => 400);
  # low quality scales:
  # to half size
  my $low = $image->scale(qtype => 'preview');
  # mixing method scale
  my $mixed = $image->scale(qtype => 'mixing', scalefactor => 0.1);
  # using an Image::Math::Constrain object
  use Image::Math::Constrain;
  my $constrain = Image::Math::Constrain->new(800, 600);
  my $scaled = $image->scale(constrain => $constrain);
  # same as Image::Math::Constrain version
  my $scaled2 = $image->scale(xpixels => 800, ypixels => 600, type => 'min');
scaleX()
scaleX() will scale along the X dimension, return a new image with the new width:

  my $newimg = $img->scaleX(pixels=>400); # 400x500
  $newimg = $img->scaleX(scalefactor=>0.25) # 175x500
  • "scalefactor" - the amount to scale the X axis. Ignored if "pixels" is provided. Default: 0.5.
  • "pixels" - the new width of the image.

Returns a new scaled image on success. The source image is not modified.

Returns false on failure, check the errstr() method for the reason for failure.

A mandatory warning is produced if scaleX() is called in void context.

scaleY()
scaleY() will scale along the Y dimension, return a new image with the new height:

  $newimg = $img->scaleY(pixels=>400); # 700x400
  $newimg = $img->scaleY(scalefactor=>0.25) # 700x125
  • "scalefactor" - the amount to scale the Y axis. Ignored if "pixels" is provided. Default: 0.5.
  • "pixels" - the new height of the image.

Returns a new scaled image on success. The source image is not modified.

Returns false on failure, check the errstr() method for the reason for failure.

A mandatory warning is produced if scaleY() is called in void context.

scale_calculate()
Performs the same calculations that the scale() method does to calculate the scaling factors from the parameters you pass.

scale_calculate() can be called as an object method, or as a class method.

Takes the following parameters over scale():

  • "width", "height" - the image width and height to base the scaling on. Required if scale_calculate() is called as a class method. If called as an object method these default to the image width and height respectively.

You might use scale_calculate() as a class method when generating an HTML "IMG" tag, for example.

Returns an empty list on failure.

Returns a list containing horizontal scale factor, vertical scale factor, new width, new height, on success.

  my ($x_scale, $y_scale, $new_width, $new_height) =
        Imager->scale_calculate(width => 1024, height => 768,
                                ypixels => 180, type => 'min');
  my ($x_scale, $y_scale, $new_width, $new_height) =
        $img->scale_calculate(xpixels => 200, type => 'min');
crop()
Another way to resize an image is to crop it. The parameters to crop are the edges of the area that you want in the returned image, where the right and bottom edges are non-inclusive. If a parameter is omitted a default is used instead.

crop() returns the cropped image and does not modify the source image.

The possible parameters are:

  • "left" - the left edge of the area to be cropped. Default: 0
  • "top" - the top edge of the area to be cropped. Default: 0
  • "right" - the right edge of the area to be cropped. Default: right edge of image.
  • "bottom" - the bottom edge of the area to be cropped. Default: bottom edge of image.
  • "width" - width of the crop area. Ignored if both "left" and "right" are supplied. Centered on the image if neither "left" nor "right" are supplied.
  • "height" - height of the crop area. Ignored if both "top" and "bottom" are supplied. Centered on the image if neither "top" nor "bottom" are supplied.

For example:

  # these produce the same image
  $newimg = $img->crop(left=>50, right=>100, top=>10, bottom=>100); 
  $newimg = $img->crop(left=>50, top=>10, width=>50, height=>90);
  $newimg = $img->crop(right=>100, bottom=>100, width=>50, height=>90);
  # and the following produce the same image
  $newimg = $img->crop(left=>50, right=>100);
  $newimg = $img->crop(left=>50, right=>100, top=>0, 
                       bottom=>$img->getheight);
  # grab the top left corner of the image
  $newimg = $img->crop(right=>50, bottom=>50);

You can also specify width and height parameters which will produce a new image cropped from the center of the input image, with the given width and height.

  $newimg = $img->crop(width=>50, height=>50);

If you supply "left", "width" and "right" values, the "right" value will be ignored. If you supply "top", "height" and "bottom" values, the "bottom" value will be ignored.

The edges of the cropped area default to the edges of the source image, for example:

  # a vertical bar from the middle from top to bottom
  $newimg = $img->crop(width=>50);
  # the right half
  $newimg = $img->crop(left=>$img->getwidth() / 2);

If the resulting image would have zero width or height then crop() returns false and $img->errstr is an appropriate error message.

A mandatory warning is produced if crop() is called in void context.

rotate()
Use the rotate() method to rotate an image. This method will return a new, rotated image.

To rotate by an exact amount in degrees or radians, use the 'degrees' or 'radians' parameter:

  my $rot20 = $img->rotate(degrees=>20);
  my $rotpi4 = $img->rotate(radians=>3.14159265/4);

Exact image rotation uses the same underlying transformation engine as the matrix_transform() method (see Imager::Engines).

You can also supply a "back" argument which acts as a background color for the areas of the image with no samples available (outside the rectangle of the source image.) This can be either an Imager::Color or Imager::Color::Float object. This is not mixed transparent pixels in the middle of the source image, it is only used for pixels where there is no corresponding pixel in the source image.

To rotate in steps of 90 degrees, use the 'right' parameter:

  my $rotated = $img->rotate(right=>270);

Rotations are clockwise for positive values.

Parameters:

  • "right" - rotate by an exact multiple of 90 degrees, specified in degrees.
  • "radians" - rotate by an angle specified in radians.
  • "degrees" - rotate by an angle specified in degrees.
  • "back" - for "radians" and "degrees" this is the color used for the areas not covered by the original image. For example, the corners of an image rotated by 45 degrees.

    This can be either an Imager::Color object, an Imager::Color::Float object or any parameter that Imager can convert to a color object, see ``Color Parameters'' in Imager::Draw for details.

    This is not mixed transparent pixels in the middle of the source image, it is only used for pixels where there is no corresponding pixel in the source image.

    Default: transparent black.

  # rotate 45 degrees clockwise, 
  my $rotated = $img->rotate(degrees => 45);
  # rotate 10 degrees counter-clockwise
  # set pixels not sourced from the original to red
  my $rotated = $img->rotate(degrees => -10, back => 'red');

Image pasting/flipping

A list of the transformations that alter the source image follows:
paste()
To copy an image to onto another image use the "paste()" method.

  $dest->paste(left=>40, top=>20, src=>$logo);

That copies the entire $logo image onto the $dest image so that the upper left corner of the $logo image is at (40,20).

Parameters:

  • "src", "img" - the source image. "src" added for compatibility with rubthrough().
  • "left", "top" - position in output of the top left of the pasted image. Default: (0,0)
  • "src_minx", "src_miny" - the top left corner in the source image to start the paste from. Default: (0, 0)
  • "src_maxx", "src_maxy" - the bottom right in the source image of the sub image to paste. This position is non inclusive. Default: bottom right corner of the source image.
  • "width", "height" - if the corresponding src_maxx or src_maxy is not defined then width or height is used for the width or height of the sub image to be pasted.

  # copy the 20x20 pixel image from (20,20) in $src_image to (10,10) in $img
  $img->paste(src=>$src_image,
              left => 10, top => 10,
              src_minx => 20, src_miny => 20,
              src_maxx => 40, src_maxx => 40);

If the source image has an alpha channel and the target doesn't, then the source is treated as if composed onto a black background.

If the source image is color and the target is gray scale, the source is treated as if run through "convert(preset=>'gray')".

rubthrough()
A more complicated way of blending images is where one image is put 'over' the other with a certain amount of opaqueness. The method that does this is rubthrough().

  $img->rubthrough(src=>$overlay,
                   tx=>30,       ty=>50,
                   src_minx=>20, src_miny=>30,
                   src_maxx=>20, src_maxy=>30);

That will take the sub image defined by $overlay and [src_minx,src_maxx)[src_miny,src_maxy) and overlay it on top of $img with the upper left corner at (30,50). You can rub 2 or 4 channel images onto a 3 channel image, or a 2 channel image onto a 1 channel image. The last channel is used as an alpha channel. To add an alpha channel to an image see convert().

Parameters:

  • "tx", "ty" - location in the target image ($self) to render the top left corner of the source.
  • "src_minx", "src_miny" - the top left corner in the source to transfer to the target image. Default: (0, 0).
  • "src_maxx", "src_maxy" - the bottom right in the source image of the sub image to overlay. This position is non inclusive. Default: bottom right corner of the source image.

  # overlay all of $source onto $targ
  $targ->rubthrough(tx => 20, ty => 25, src => $source);
  # overlay the top left corner of $source onto $targ
  $targ->rubthrough(tx => 20, ty => 25, src => $source,
                    src_maxx => 20, src_maxy => 20);
  # overlay the bottom right corner of $source onto $targ
  $targ->rubthrough(tx => 20, ty => 30, src => $src,
                    src_minx => $src->getwidth() - 20,
                    src_miny => $src->getheight() - 20);

rubthrough() returns true on success. On failure check "$target->errstr" for the reason for failure.

compose()
Draws the source image over the target image, with the source alpha channel modified by the optional mask and the opacity.

  $img->compose(src=>$overlay,
                tx=>30,       ty=>50,
                src_minx=>20, src_miny=>30,
                src_maxx=>20, src_maxy=>30,
                mask => $mask, opacity => 0.5);

That will take the sub image defined by $overlay and [src_minx,src_maxx)[src_miny,src_maxy) and overlay it on top of $img with the upper left corner at (30,50). You can rub 2 or 4 channel images onto a 3 channel image, or a 2 channel image onto a 1 channel image.

Parameters:

  • "src" - the source image to draw onto the target. Required.
  • "tx", "ty" - location in the target image ($self) to render the top left corner of the source. These can also be supplied as "left" and "right". Default: (0, 0).
  • "src_minx", "src_miny" - the top left corner in the source to transfer to the target image. Default: (0, 0).
  • "src_maxx", "src_maxy" - the bottom right in the source image of the sub image to overlay. This position is non inclusive. Default: bottom right corner of the source image.
  • "mask" - a mask image. The first channel of this image is used to modify the alpha channel of the source image. This can be used to mask out portions of the source image. Where the first channel is zero none of the source image will be used, where the first channel is maximum the full alpha of the source image will be used, as further modified by the opacity.
  • opacity - further modifies the alpha channel of the source image, in the range 0.0 to 1.0. Default: 1.0.
  • combine - the method to combine the source pixels with the target. See the combine option documentation in Imager::Fill. Default: normal.

Calling compose() with no mask, combine set to "normal", opacity set to 1.0 is equivalent to calling rubthrough().

compose() is intended to be produce similar effects to layers in interactive paint software.

  # overlay all of $source onto $targ
  $targ->compose(tx => 20, ty => 25, src => $source);
  # overlay the top left corner of $source onto $targ
  $targ->compose(tx => 20, ty => 25, src => $source,
                    src_maxx => 20, src_maxy => 20);
  # overlay the bottom right corner of $source onto $targ
  $targ->compose(tx => 20, ty => 30, src => $src,
                    src_minx => $src->getwidth() - 20,
                    src_miny => $src->getheight() - 20);

compose() returns true on success. On failure check $target->errstr for the reason for failure.

flip()
An inplace horizontal or vertical flip is possible by calling the "flip()" method. If the original is to be preserved it's possible to make a copy first. The only parameter it takes is the "dir" parameter which can take the values "h", "v", "vh" and "hv".

  $img->flip(dir=>"h");       # horizontal flip
  $img->flip(dir=>"vh");      # vertical and horizontal flip
  $nimg = $img->copy->flip(dir=>"v"); # make a copy and flip it vertically

flip() returns true on success. On failure check $img->errstr for the reason for failure.

Color transformations

convert()
You can use the convert method to transform the color space of an image using a matrix. For ease of use some presets are provided.

The convert method can be used to:

  • convert an RGB or RGBA image to gray scale.
  • convert a gray scale image to RGB.
  • extract a single channel from an image.
  • set a given channel to a particular value (or from another channel)

The currently defined presets are:

  • "gray", "grey" - converts an RGBA image into a gray scale image with alpha channel, or an RGB image into a gray scale image without an alpha channel.

    This weights the RGB channels at 22.2%, 70.7% and 7.1% respectively.

  • "noalpha" - removes the alpha channel from a 2 or 4 channel image. An identity for other images.
  • "red", "channel0" - extracts the first channel of the image into a single channel image
  • "green", "channel1" - extracts the second channel of the image into a single channel image
  • "blue", "channel2" - extracts the third channel of the image into a single channel image
  • "alpha" - extracts the alpha channel of the image into a single channel image.

    If the image has 1 or 3 channels (assumed to be gray scale or RGB) then the resulting image will be all white.

  • "rgb"

    converts a gray scale image to RGB, preserving the alpha channel if any

  • "addalpha" - adds an alpha channel to a gray scale or RGB image. Preserves an existing alpha channel for a 2 or 4 channel image.

For example, to convert an RGB image into a gray scale image:

  $new = $img->convert(preset=>'grey'); # or gray

or to convert a gray scale image to an RGB image:

  $new = $img->convert(preset=>'rgb');

The presets aren't necessary simple constants in the code, some are generated based on the number of channels in the input image.

If you want to perform some other color transformation, you can use the 'matrix' parameter.

For each output pixel the following matrix multiplication is done:

  | channel[0] |   | $c00, ...,  $c0k |   | inchannel[0] |
  |    ...     | = |       ...        | x |     ...      |
  | channel[k] |   | $ck0, ...,  $ckk |   | inchannel[k] |
                                                          1
Where C<k = $img-E<gt>getchannels()-1>.

So if you want to swap the red and green channels on a 3 channel image:

  $new = $img->convert(matrix=>[ [ 0, 1, 0 ],
                                 [ 1, 0, 0 ],
                                 [ 0, 0, 1 ] ]);

or to convert a 3 channel image to gray scale using equal weightings:

  $new = $img->convert(matrix=>[ [ 0.333, 0.333, 0.334 ] ])

Convert a 2 channel image (gray scale with alpha) to an RGBA image with the gray converted to the specified RGB color:

  # set (RGB) scaled on the grey scale portion and copy the alpha
  # channel as is
  my $colored = $gray->convert(matrix=>[ [ ($red/255),   0 ], 
                                         [ ($green/255), 0 ], 
                                         [ ($blue/255),  0 ], 
                                         [ 0,            1 ],
                                       ]);

To convert a 3 channel image to a 4 channel image with a 50 percent alpha channel:

  my $withalpha = $rgb->convert(matrix =>[ [ 1, 0, 0, 0 ],
                                           [ 0, 1, 0, 0 ],
                                           [ 0, 0, 1, 0 ],
                                           [ 0, 0, 0, 0.5 ],
                                         ]);
combine()
Combine channels from one or more input images into a new image.

Parameters:

  • "src" - a reference to an array of input images. There must be at least one input image. A given image may appear more than once in "src".
  • "channels" - a reference to an array of channels corresponding to the source images. If "channels" is not supplied then the first channel from each input image is used. If the array referenced by "channels" is shorter than that referenced by "src" then the first channel is used from the extra images.

  # make an rgb image from red, green, and blue images
  my $rgb = Imager->combine(src => [ $red, $green, $blue ]);
  # convert a BGR image into RGB
  my $rgb = Imager->combine(src => [ $bgr, $bgr, $bgr ],
                            channels => [ 2, 1, 0 ]);
  # add an alpha channel from another image
  my $rgba = Imager->combine(src => [ $rgb, $rgb, $rgb, $alpha ],
                     channels => [ 0, 1, 2, 0 ]);

Color Mappings

map()
You can use the map method to map the values of each channel of an image independently using a list of look-up tables. It's important to realize that the modification is made inplace. The function simply returns the input image again or undef on failure.

Each channel is mapped independently through a look-up table with 256 entries. The elements in the table should not be less than 0 and not greater than 255. If they are out of the 0..255 range they are clamped to the range. If a table does not contain 256 entries it is silently ignored.

Single channels can mapped by specifying their name and the mapping table. The channel names are "red", "green", "blue", "alpha".

  @map = map { int( $_/2 } 0..255;
  $img->map( red=>\@map );

It is also possible to specify a single map that is applied to all channels, alpha channel included. For example this applies a gamma correction with a gamma of 1.4 to the input image.

  $gamma = 1.4;
  @map = map { int( 0.5 + 255*($_/255)**$gamma ) } 0..255;
  $img->map(all=> \@map);

The "all" map is used as a default channel, if no other map is specified for a channel then the "all" map is used instead. If we had not wanted to apply gamma to the alpha channel we would have used:

  $img->map(all=> \@map, alpha=>[]);

Since "[]" contains fewer than 256 element the gamma channel is unaffected.

It is also possible to simply specify an array of maps that are applied to the images in the RGBA order. For example to apply maps to the "red" and "blue" channels one would use:

  $img->map(maps=>[\@redmap, [], \@bluemap]);

AUTHOR

Tony Cook <[email protected]>, Arnar M. Hrafnkelsson

REVISION

$Revision$