NAME
Prima::Image - Bitmap routines
SYNOPSIS
use Prima qw(Application);
# create a new image from scratch
my $i = Prima::Image-> new(
width => 32,
height => 32,
type => im::BW, # same as im::bpp1 | im::GrayScale
);
# draw something
$i-> begin_paint;
$i-> color( cl::White);
$i-> ellipse( 5, 5, 10, 10);
$i-> end_paint;
# mangle
$i-> size( 64, 64);
# file operations
$i-> save('a.gif') or die "Error saving:$@\n";
$i-> load('a.gif') or die "Error loading:$@\n";
# draw on screen
$::application-> begin_paint;
# an image is drawn as specified by its palette
$::application-> set( color => cl::Red, backColor => cl::Green);
$::application-> put_image( 100, 100, $i);
# a bitmap is drawn as specified by destination device colors
$::application-> put_image( 200, 100, $i-> bitmap);
DESCRIPTION
Prima::Image, Prima::Icon and Prima::DeviceBitmap are classes for
bitmap handling, including file and graphic input and output.
Prima::Image and Prima::DeviceBitmap are descendants of Prima::Drawable
and represent bitmaps, stored in memory. Prima::Icon is a descendant
of Prima::Image and contains a transparency mask along with the regular
data.
USAGE
Images usually are represented as a memory area, where pixel data are
stored row-wise. The Prima toolkit is no exception, however, it does
not assume that the GUI system uses the same memory format. The
implicit conversion routines are called when Prima::Image is about to
be drawn onto the screen, for example. The conversions are not always
efficient, therefore the Prima::DeviceBitmap class is introduced to
represent a bitmap, stored in the system memory in the system pixel
format. These two basic classes serve the different needs, but can be
easily converted to each other, with "image" and "bitmap" methods.
Prima::Image is a more general bitmap representation, capable of file
and graphic input and output, plus it is supplied with number of
conversion and scaling functions. The Prima::DeviceBitmap class has
almost none of additional functionality, and is targeted to efficient
graphic input and output.
Graphic input and output
As descendants of Prima::Drawable, all Prima::Image, Prima::Icon and
Prima::DeviceBitmap objects are subject to three-state painting mode -
normal ( disabled ), painting ( enabled ) and informational.
Prima::DeviceBitmap is, however, exists only in the enabled state, and
can not be switched to the other two.
When an object enters the enabled state, it serves as a canvas, and all
Prima::Drawable operations can be performed on it. When the object is
back to the disabled state, the graphic information is stored into the
object associated memory, in the pixel format, supported by the
toolkit. This information can be visualized by using one of
"Prima::Drawable::put_image" group methods. If the object enters the
enabled state again, the graphic information is presented as an initial
state of a bitmap.
It must be noted, that if an implicit conversion takes place after an
object enters and before it leaves the enabled state, as it is with
Prima::Image and Prima::Icon, the bitmap is converted to the system
pixel format. During such conversion some information can be lost, due
to down-sampling, and there is no way to preserve the information. This
does not happen with Prima::DeviceBitmap.
Image objects can be drawn upon images, as well as on the screen and
Prima::Widget objects. This operation is performed via one of
Prima::Drawable::put_image group methods ( see Prima::Drawable), and
can be called with the image object disregarding the paint state. The
following code illustrates the dualism of an image object, where it can
serve both as a drawing surface and as a drawing tool:
my $a = Prima::Image-> create( width => 100, height => 100, type => im::RGB);
$a-> begin_paint;
$a-> clear;
$a-> color( cl::Green);
$a-> fill_ellipse( 50, 50, 30, 30);
$a-> end_paint;
$a-> rop( rop::XorPut);
$a-> put_image( 10, 10, $a);
$::application-> begin_paint;
$::application-> put_image( 0, 0, $a);
$::application-> end_paint;
It must be noted, that "put_image", "stretch_image" and
"put_image_indirect" are only painting methods that allow drawing on an
image that is in its paint-disabled state. Moreover, in such context
they only allow "Prima::Image" descendants to be passed as a source
image object. This functionality does not imply that the image is
internally switched to the paint-enabled state and back; the painting
is performed without switching and without interference with the
system’s graphical layer.
Another special case is a 1-bit ( monochrome ) DeviceBitmap. When it is
drawn upon a drawable with bit depth greater than 1, the drawable’s
color and backColor properties are used to reflect 1 and 0 bits,
respectively. On a 1-bit drawable this does not happen, and the color
properties are not used.
File input and output
Depending on the toolkit configuration, images can be read and written
in different formats. This functionality in accessible via "load()" and
"save()" methods. Prima::image-load is dedicated to the description of
loading and saving parameters, that can be passed to the methods, so
they can handle different aspects of file format-specific options, such
as multi-frame operations, auto conversion when a format does not
support a particular pixel format etc. In this document, "load()" and
"save()" methods are illustrated only in their basic, single-frame
functionality. When called with no extra parameters, these methods fail
only if a disk I/O error occurred or an unknown image format was used.
When an image is loaded, the old bitmap memory content is discarded,
and the image attributes are changed accordingly to the loaded image.
Along with these, an image palette is loaded, if available, and a pixel
format is assigned, closest or identical to the pixel format in the
image file.
Pixel formats
Prima::Image supports a number of pixel formats, governed by the
"::type" property. It is reflected by an integer value, a combination
of "im::XXX" constants. The whole set of pixel formats is represented
by colored formats, like, 16-color, 256-color and 16M-color, and by
gray-scale formats, mapped to C data types - unsigned char, unsigned
short, unsigned long, float and double. The gray-scale formats are
further subdivided to real-number formats and complex-number format;
the last ones are represented by two real values per pixel, containing
the real and the imaginary values.
Prima::Image can also be initialized from other formats, that it does
not support, but can convert data from. Currently these are represented
by a set of permutations of 32-bit RGBA format, and 24-bit BGR format.
These formats can only be used in conjunction with "::data" property.
The conversions can be performed between any of the supported formats (
to do so, "::type" property is to be set-called ). An image of any of
these formats can be drawn on the screen, but if the system can not
accept the pixel format ( as it is with non-integer or complex formats
), the bitmap data are implicitly converted. The conversion does not
change the data if the image is about to be drawn; the conversion is
performed only when the image is about to be served as a drawing
surface. If, by any reason, it is desired that the pixel format is not
to be changed, the "::preserveType" property must be set to 1. It does
not prevent the conversion, but it detects if the image was implicitly
converted inside "end_paint()" call, and reverts it to its previous
pixel format.
There are situations, when pixel format must be changed together while
down-sampling the image. One of four down-sampling methods can be
selected - normal, 8x8 ordered halftoning, error diffusion, and error
diffusion combined with optimized palette. These can be set to the
"::conversion" property with one of "ict::XXX" constants. When there
is no information loss, "::conversion" property is not used.
Another special case of conversion is a conversion with a palette. The
following calls,
$image-> type( im::bpp4);
$image-> palette( $palette);
and
$image-> palette( $palette);
$image-> type( im::bpp4);
produce different results, but none of these takes into account
eventual palette remapping, because "::palette" property does not
change bitmap pixel data, but overwrites palette information. A proper
call syntax here would be
$image-> set(
palette => $palette,
type => im::bpp4,
);
This call produces also palette pixel mapping. This syntax is most
powerful when conversion is set to "ict::Optimized" ( by default ). It
not only allows remapping or downsampling to a predefined colors set,
but also can be used to limit palette size to a particular number,
without knowing the actual values of the final color palette. For
example, for an 24-bit image,
$image-> set( type => im::bpp8, palette => 32);
call would calculate colors in the image, compress them to an optimized
palette of 32 cells and finally converts to a 8-bit format.
Instead of "palette" property, "colormap" can also be used.
Data access
The pixel values can be accessed in Prima::Drawable style, via
"::pixel" property. However, Prima::Image introduces several helper
functions, for different aims. The "::data" property is used to set or
retrieve a scalar representation of bitmap data. The data are expected
to be lined up to a ’line size’ margin ( 4-byte boundary ), which is
calculated as
$lineSize = int(( $image->width * ( $image-> type & im::BPP) + 31) / 32) * 4;
or returned from the read-only property "::lineSize".
This is the line size for the data as lined up internally in memory,
however "::data" should not necessarily should be aligned like this,
and can be accompanied with a write-only flag ’lineSize’ if pixels are
aligned differently:
$image-> set( width => 1, height=> 2);
$image-> type( im::RGB);
$image-> set(
data => 'RGB----RGB----',
lineSize => 7,
);
print $image-> data, "\n";
output: RGB-RGB-
Internally, Prima contains images in memory so that the first scanline
is the farthest away from the memory start; this is consistent with
general Y-axis orientation in Prima drawable terminology, but might be
inconvenient when importing data organized otherwise. Another write-
only boolean flag "reverse" can be set to 1 so data then are treated as
if the first scanline of the image is the closest to the start of data:
$image-> set( width => 1, height=> 2, type => im::RGB);
$image-> set(
data => 'RGB-123-',
reverse => 1,
);
print $image-> data, "\n";
output: RGB-123-
Although it is possible to perform all kinds of calculations and
modification with the pixels, returned by "::data", it is not advisable
unless the speed does not matter. Standalone PDL package with help of
PDL::PrimaImage package, and Prima-derived IPA package provide routines
for data and image analysis. Also, Prima::Image::Magick connects
ImageMagick with Prima. Prima::Image itself provides only the simplest
statistic information, namely: lowest and highest pixel values, pixel
sum, sum of square pixels, mean, variance, and standard deviation.
Standalone usage
The image functionality can be used standalone, with all other parts of
the toolkit being uninitialized. This is useful in non-interactive
programs, running in evnironments with no GUI access, a cgi-script with
no access to X11 display, for example. Normally, Prima fails to start
in such situations, but can be told not to initialize its GUI part by
explicitly operating system-dependent options. To do so, invoke
use Prima::noX11;
in the beginning of your program. See Prima::noX11 for more.
Prima::Icon
Prima::Icon inherits all properties of Prima::Image, and it also
provides a 1-bit depth transparency mask. This mask can also be loaded
and saved into image files, if the format supports a transparency
information.
Similar to Prima::Image::data property, Prima::Icon::mask property
provides access to the binary mask data. The mask can be updated
automatically, after an icon object was subject to painting, resizing,
or other destructive change. The auxiliary properties "::autoMasking"
and "::maskColor"/"::maskIndex" regulate mask update procedure. For
example, if an icon was loaded with the color ( vs. bitmap )
transparency information, the binary mask will be generated anyway, but
it will be also recorded that a particular color serves as a
transparent indicator, so eventual conversions can rely on the color
value, instead of the mask bitmap.
If an icon is drawn upon a graphic canvas, the image output is
constrained to the mask. On raster displays it is typically simulated
by a combination of and- and xor- operation modes, therefore attempts
to put an icon with "::rop", different from "rop::CopyPut", usually
fail.
API
Prima::Image properties
colormap @PALETTE
A color palette, used for representing 1, 4, and 8-bit bitmaps,
when an image object is to be visualized. @PALETTE contains
individual colors component triplets, in RGB format. For example,
black-and-white monochrome image may contain colormap as
"0,0xffffff".
See also "palette".
conversion TYPE
Selects the type of dithering algorithm to be used for pixel down-
sampling. TYPE is one of "ict::XXX" constants:
ict::None - no dithering
ict::Halftone - 8x8 ordered halftone dithering
ict::ErrorDiffusion - error diffusion dithering with static palette
ict::Optimized - error diffusion dithering with optimized palette
As an example, if a 4x4 color image with every pixel set to
RGB(32,32,32), converted to a 1-bit image, the following results
occur:
ict::None:
[ 0 0 0 0 ]
[ 0 0 0 0 ]
[ 0 0 0 0 ]
[ 0 0 0 0 ]
ict::Halftone:
[ 0 0 0 0 ]
[ 0 0 1 0 ]
[ 0 0 0 0 ]
[ 1 0 0 0 ]
ict::ErrorDiffusion, ict::Ordered:
[ 0 0 1 0 ]
[ 0 0 0 1 ]
[ 0 0 0 0 ]
[ 0 0 0 0 ]
data SCALAR
Provides access to the bitmap data. On get-call, returns all bitmap
pixels, aligned to 4-byte boundary. On set-call, stores the
provided data with same alignment. The alignment can be altered by
submitting ’lineSize’ write-only flag to set call; the ordering of
scan lines can be altered by setting ’reverse’ write-only flag (
see "Data access" ).
height INTEGER
Manages the vertical dimension of the image data. On set-call, the
image data are changed accordingly to the new height, and depending
on "::vScaling" property, the pixel values are either scaled or
truncated.
hScaling BOOLEAN
If 1, the bitmap data will be scaled when image changes its
horizontal extent. If 0, the data will be stripped or padded with
zeros.
lineSize INTEGER
A read-only property, returning the length of an image row in
bytes, as represented internally in memory. Data returned by
"::data" property are aligned with "::lineSize" bytes per row, and
setting "::data" expects data aligned with this value, unless
"lineSize" is set together with "data" to indicate another
alignment. See "Data access" for more.
mean
Returns mean value of pixels. Mean value is "::sum" of pixel
values, divided by number of pixels.
palette [ @PALETTE ]
A color palette, used for representing 1, 4, and 8-bit bitmaps,
when an image object is to be visualized. @PALETTE contains
individual color component triplets, in BGR format. For example,
black-and-white monochrome image may contain palette as
"[0,0,0,255,255,255]".
See also "colormap".
pixel ( X_OFFSET, Y_OFFSET ) PIXEL
Provides per-pixel access to the image data when image object is in
disabled paint state. Otherwise, same as "Prima::Drawable::pixel".
preserveType BOOLEAN
If 1, reverts the image type to its old value if an implicit
conversion was called during "end_paint()".
rangeHi
Returns maximum pixel value in the image data.
rangeLo
Returns minimum pixel value in the image data.
size WIDTH, HEIGHT
Manages dimensions of the image. On set-call, the image data are
changed accordingly to the new dimensions, and depending on
"::vScaling" and "::hScaling" properties, the pixel values are
either scaled or truncated.
stats ( INDEX ) VALUE
Returns one of calculated values, that correspond to INDEX, which
is one of the following "is::XXX" constants:
is::RangeLo - minimum pixel value
is::RangeHi - maximum pixel value
is::Mean - mean value
is::Variance - variance
is::StdDev - standard deviation
is::Sum - sum of pixel values
is::Sum2 - sum of squares of pixel values
The values are re-calculated on request and cached. On set-call
VALUE is stored in the cache, and is returned on next get-call.
The cached values are discarded every time the image data changes.
These values are also accessible via set of alias properties:
"::rangeLo", "::rangeHi", "::mean", "::variance", "::stdDev",
"::sum", "::sum2".
stdDev
Returns standard deviation of the image data. Standard deviation
is the square root of "::variance".
sum Returns sum of pixel values of the image data
sum2
Returns sum of squares of pixel values of the image data
type TYPE
Governs the image pixel format type. TYPE is a combination of
"im::XXX" constants. The constants are collected in groups:
Bit-depth constants provide size of pixel is bits. Their actual
value is same as number of bits, so "im::bpp1" value is 1,
"im::bpp4" - 4, etc. The valid constants represent bit depths from
1 to 128:
im::bpp1
im::bpp4
im::bpp8
im::bpp16
im::bpp24
im::bpp32
im::bpp64
im::bpp128
The following values designate the pixel format category:
im::Color
im::GrayScale
im::RealNumber
im::ComplexNumber
im::TrigComplexNumber
Value of "im::Color" is 0, whereas other category constants
represented by unique bit value, so combination of "im::RealNumber"
and "im::ComplexNumber" is possible.
There also several mnemonic constants defined:
im::Mono - im::bpp1
im::BW - im::bpp1 | im::GrayScale
im::16 - im::bpp4
im::Nibble - im::bpp4
im::256 - im::bpp8
im::RGB - im::bpp24
im::Triple - im::bpp24
im::Byte - gray 8-bit unsigned integer
im::Short - gray 16-bit unsigned integer
im::Long - gray 32-bit unsigned integer
im::Float - float
im::Double - double
im::Complex - dual float
im::DComplex - dual double
im::TrigComplex - dual float
im::TrigDComplex - dual double
Bit depths of float- and double- derived pixel formats depend on a
platform.
The groups can be masked out with the mask values:
im::BPP - bit depth constants
im::Category - category constants
im::FMT - extra format constants
The extra formats are the pixel formats, not supported by "::type",
but recognized within the combined set-call, like
$image-> set(
type => im::fmtBGRI,
data => 'BGR-BGR-',
);
The data, supplied with the extra image format specification will
be converted to the closest supported format. Currently, the
following extra pixel formats are recognized:
im::fmtBGR
im::fmtRGBI
im::fmtIRGB
im::fmtBGRI
im::fmtIBGR
variance
Returns variance of pixel values of the image data. Variance is
"::sum2", divided by number of pixels minus square of "::sum" of
pixel values.
vScaling BOOLEAN
If 1, the bitmap data will be scaled when image changes its
vertical extent. If 0, the data will be stripped or padded with
zeros.
width INTEGER
Manages the horizontal dimension of the image data. On set-call,
the image data are changed accordingly to the new width, and
depending on "::hScaling" property, the pixel values are either
scaled or truncated.
Prima::Icon properties
autoMasking TYPE
Selects whether the mask information should be updated
automatically with "::data" change or not. Every "::data" change is
mirrored in "::mask", using TYPE, one of "am::XXX" constants:
am::None - no mask update performed
am::MaskColor - mask update based on ::maskColor property
am::MaskIndex - mask update based on ::maskIndex property
am::Auto - mask update based on corner pixel values
The "::maskColor" color value is used as a transparent color if
TYPE is "am::MaskColor". The transparency mask generation
algorithm, turned on by "am::Auto" checks corner pixel values,
assuming that majority of the corner pixels represents a
transparent color. Once such color is found, the mask is generated
as in "am::MaskColor" case.
"::maskIndex" is the same as "::maskColor", except that it points
to a specific color index in the palette.
When image "::data" is stretched, "::mask" is stretched
accordingly, disregarding the "::autoMasking" value.
mask SCALAR
Provides access to the transparency bitmap. On get-call, returns
all bitmap pixels, aligned to 4-byte boundary in 1-bit format. On
set-call, stores the provided transparency data with same
alignment.
maskColor COLOR
When "::autoMasking" set to "am::MaskColor", COLOR is used as a
transparency value.
maskIndex INDEX
When "::autoMasking" set to "am::MaskIndex", INDEXth color in teh
current palette is used as a transparency value.
Prima::DeviceBitmap properties
monochrome BOOLEAN
A read-only property, that can only be set during creation,
reflects whether the system bitmap is black-and-white 1-bit
(monochrome) or not. The color depth of a bitmap can be read via
"get_bpp()" method; monochrome bitmaps always have bit depth of 1.
Prima::Image methods
bitmap
Returns newly created Prima::DeviceBitmap instance, with the image
dimensions and with the bitmap pixel values copied to.
codecs
Returns array of hashes, each describing the supported image
format. If the array is empty, the toolkit was set up so it can not
load and save images.
See Prima::image-load for details.
This method can be called without object instance.
dup Returns a duplicate of the object, a newly created Prima::Image,
with all information copied to it.
extract X_OFFSET, Y_OFFSET, WIDTH, HEIGHT
Returns a newly created image object with WIDTH and HEIGHT
dimensions, initialized with pixel data from X_OFFSET and Y_OFFSET
in the bitmap.
get_bpp
Returns the bit depth of the pixel format. Same as "::type &
im::BPP".
get_handle
Returns a system handle for an image object.
load (FILENAME or FILEGLOB) [ %PARAMETERS ]
Loads image from file FILENAME or stream FILEGLOB into an object,
and returns the success flag. The semantics of "load()" is
extensive, and can be influenced by PARAMETERS hash. "load()" can
be called either in a context of an existing object, then a boolean
success flag is returned, or in a class context, then a newly
created object ( or "undef" ) is returned. If an error occurs, $@
variable contains the error description string. These two
invocation semantics are equivalent:
my $x = Prima::Image-> create();
die "$@" unless $x-> load( ... );
and
my $x = Prima::Image-> load( ... );
die "$@" unless $x;
See Prima::image-load for details.
NB! When loading from streams on win32, mind "binmode".
map COLOR
Performs iterative mapping of bitmap pixels, setting every pixel to
"::color" property with respect to "::rop" type if a pixel equals
to COLOR, and to "::backColor" property with respect to "::rop2"
type otherwise.
"rop::NoOper" type can be used for color masking.
Examples:
width => 4, height => 1, data => [ 1, 2, 3, 4]
color => 10, backColor => 20, rop => rop::CopyPut
rop2 => rop::CopyPut
input: map(2) output: [ 20, 10, 20, 20 ]
rop2 => rop::NoOper
input: map(2) output: [ 1, 10, 3, 4 ]
resample SRC_LOW, SRC_HIGH, DEST_LOW, DEST_HIGH
Performs linear scaling of gray pixel values from range (SRC_LOW -
SRC_HIGH) to range (DEST_LOW - DEST_HIGH). Can be used to visualize
gray non-8 bit pixel values, by the code:
$image-> resample( $image-> rangeLo, $image-> rangeHi, 0, 255);
save (FILENAME or FILEGLOB), [ %PARAMETERS ]
Stores image data into image file FILENAME or stream FILEGLOB, and
returns the success flag. The semantics of "save()" is extensive,
and can be influenced by PARAMETERS hash. If error occurs, $@
variable contains error description string.
Note that when saving to a stream, "codecID" must be explicitly
given in %PARAMETERS.
See Prima::image-load for details.
NB! When saving to streams on win32, mind "binmode".
Prima::Image events
"Prima::Image"-specific events occur only from inside load call, to
report image loading progress. Not all codecs (currently JPEG,PNG,TIFF
only) are able to report the progress to the caller. See "Loading with
progress indicator" in Prima::image-load for details,
"watch_load_progress" in Prima::ImageViewer and "load" in
Prima::ImageDialog for suggested use.
HeaderReady
Called whenever image header is read, and image dimensions and
pixel type is changed accordingly to accomodate image data.
DataReady X, Y, WIDTH, HEIGHT
Called whenever image data that cover area designated by
X,Y,WIDTH,HEIGHT is acquired. Use "load" option "eventDelay" to
limit the rate of "DataReady" event.
Prima::Icon methods
split
Returns two new Prima::Image objects of same dimension. Pixels in
the first is are duplicated from "::data" storage, in the second -
from "::mask" storage.
combine DATA, MASK
Copies information from DATA and MASK images into "::data" and
"::mask" property. DATA and MASK are expected to be images of same
dimension.
Prima::DeviceBitmap methods
icon
Returns a newly created Prima::Icon object instance, with the pixel
information copied from the object.
image
Returns a newly created Prima::Image object instance, with the
pixel information copied from the object.
get_handle
Returns a system handle for a system bitmap object.
AUTHOR
Dmitry Karasik, <dmitry@karasik.eu.org>.
SEE ALSO
Prima, Prima::Drawable, Prima::image-load, Prima::codecs.
PDL, PDL::PrimaImage, IPA
ImageMagick, Prima::Image::Magick