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.
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.
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.
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