[imager.git] / lib / Imager / Transformations.pod

=head1 NAME

Imager::Transformations - Simple transformations of one image into another.

=head1 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->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 ] ]);

  # 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

=head1 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 C<flip()>, C<paste()> and
C<rubthrough()>.  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');

=head2 Image copying/resizing/cropping/rotating

A list of the transformations that do not alter the source image follows:

=over

=item copy

To create a copy of an image use the C<copy()> method.  This is usefull
if you want to keep an original after doing something that changes the image.

  $newimg = $orig->copy();

=item scale

X<scale>To scale an image so porportions are maintained use the
C<$img-E<gt>scale()> method.  if you give either a xpixels or ypixels
parameter they will determine the width or height respectively.  If
both are given the one resulting in a larger image is used, unless you
set the C<type> parameter to C<'min'>.  example: C<$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
C<qtype=E<gt>'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.

=over

=item *

xpixels, ypixels - desired size of the scaled image.  The C<type>
parameter controls whether the larger or smaller of the two possible
sizes is chosen, or if the image is scaled non-proportionally.

=item *

constrain - an Image::Math::Constrain object defining the way in which
the image size should be constrained.

=item *

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.

=item *

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.

=item *

type - controls whether the larger or smaller of the two possible
sizes is chosen, possible values are:

=over

=item *

min - the smaller of the 2 sizes are chosen.

=item *

max - the larger of the 2 sizes.  This is the default.

=item *

nonprop - non-proportional scaling.  New in Imager 0.54.

=back

scale() will fail if C<type> is set to some other value.

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

C<type> is only used if both C<xpixels> and C<ypixels> are supplied.

=item *

qtype - defines the quality of scaling performed.  Possible values are:

=over

=item *

C<normal> - high quality scaling.  This is the default.

=item *

C<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.

=item *

C<mixing> - implements the mixing algorithm implemented by pnmscale.
This retains more detail when scaling down than C<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.

=back

scale() will fail if C<qtype> is set to some other value.

C<preview> is faster than C<mixing> which is much faster than C<normal>.

=back

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 C<xpixels> and C<ypixels> and setting C<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');

=item 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

=over

=item *

scalefactor - the amount to scale the X axis.  Ignored if C<pixels> is
provided.  Default: 0.5.

=item *

pixels - the new width of the image.

=back

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.

=item 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

=over

=item *

scalefactor - the amount to scale the Y axis.  Ignored if C<pixels> is
provided.  Default: 0.5.

=item *

pixels - the new height of the image.

=back

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.

=item 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:

=over

=item *

C<left> - the left edge of the area to be cropped.  Default: 0

=item *

C<top> - the top edge of the area to be cropped.  Default: 0

=item *

C<right> - the right edge of the area to be cropped.  Default: right
edge of image.

=item *

C<bottom> - the bottom edge of the area to be cropped.  Default:
bottom edge of image.

=item *

C<width> - width of the crop area.  Ignored if both C<left> and C<right> are
supplied.  Centered on the image if neither C<left> nor C<right> are
supplied.

=item *

C<height> - height of the crop area.  Ignored if both C<top> and
C<bottom> are supplied.  Centered on the image if neither C<top> nor
C<bottom> are supplied.

=back

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

=item 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 C<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 B<not> mixed
transparent pixels in the middle of the source image, it is B<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:

=over

=item *

right - rotate by an exact multiple of 90 degrees, specified in
degreess.

=item *

radians - rotate by an angle specified in radians.

=item *

degrees - rotate by an angle specified in degrees.

=item *

back - for C<radians> and C<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
L<Imager::Draw/Color Parameters> for details.

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

Default: transparent black.

=back

  # 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');

=back

=head2 Image pasting/flipping

A list of the transformations that alter the source image follows:

=over

=item paste

X<paste>To copy an image to onto another image use the C<paste()>
method.

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

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

Parameters:

=over

=item *

src, img - the source image.  I<src> added for compatibility with
rubthrough().

=item *

left, top - position in output of the top left of the pasted image.
Default: (0,0)

=item *

src_minx, src_miny - the top left corner in the source image to start
the paste from.  Default: (0, 0)

=item *

src_maxx, src_maxy - the bottom right in the source image of the sub
image to paste.  This position is B<non> inclusive.  Default: bottom
right corner of the source image.

=item *

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.

=back

  # 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);
              
=item 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 I<$overlay> and
I<[src_minx,src_maxx)[src_miny,src_maxy)> and overlay it on top of
I<$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 I<convert()>.

Parameters:

=over

=item *

tx, ty - location in the the target image ($self) to render the top
left corner of the source.

=item *

src_minx, src_miny - the top left corner in the source to transfer to
the target image.  Default: (0, 0).

=item *

src_maxx, src_maxy - the bottom right in the source image of the sub
image to overlay.  This position is B<non> inclusive.  Default: bottom
right corner of the source image.

=back

  # 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.

=item flip

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

=back

=head2 Color transformations

=over

=item 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:

=over

=item *

convert an RGB or RGBA image to grayscale.

=item *

convert a grayscale image to RGB.

=item *

extract a single channel from an image.

=item *

set a given channel to a particular value (or from another channel)

=back

The currently defined presets are:

=over

=item gray

=item grey

converts an RGBA image into a grayscale image with alpha channel, or
an RGB image into a grayscale image without an alpha channel.

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

=item noalpha

removes the alpha channel from a 2 or 4 channel image.  An identity
for other images.

=item red

=item channel0

extracts the first channel of the image into a single channel image

=item green

=item channel1

extracts the second channel of the image into a single channel image

=item blue

=item channel2

extracts the third channel of the image into a single channel image

=item alpha

extracts the alpha channel of the image into a single channel image.

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

=item rgb

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

=item addalpha

adds an alpha channel to a grayscale or RGB image.  Preserves an
existing alpha channel for a 2 or 4 channel image.

=back

For example, to convert an RGB image into a greyscale image:

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

or to convert a grayscale 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 colour 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 greyscale using equal weightings:

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

Convert a 2 channel image (grayscale with alpha) to an RGBA image with
the grey 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 ],
                                         ]);

=back

=head2 Color Mappings

=over

=item map

You can use the map method to map the values of each channel of an
image independently using a list of lookup 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 lookup 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 C<red>, C<green>, C<blue>, C<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 C<all> map is used as a default channel, if no other map is
specified for a channel then the C<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 C<[]> 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 C<red> and C<blue> channels one would use:

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

=back

=head1 SEE ALSO

L<Imager>, L<Imager::Engines>

=head1 AUTHOR

Tony Cook <tony@imager.perl.org>, Arnar M. Hrafnkelsson

=head1 REVISION

$Revision$

=cut