Prima::Image - Bitmap routines |
Prima::Image - Bitmap routines
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);
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.
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.
Note: If you're looking for information how to display an image, this is not
the manual page. Look either at the Prima::ImageViewer manpage, or use put_image
/
stretch_image
( the Prima::Drawable manpage ) inside your widget's onPaint.
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 the Prima::Widget manpage objects. This operation is performed via one of Prima::Drawable::put_image group methods ( see the Prima::Drawable manpage), 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. the Prima::image-load manpage 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.
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 the PDL::PrimaImage manpage
package, and Prima-derived IPA package provide routines for data and image
analysis. Also, the Prima::Image::Magick manpage 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.
Some of image functionality can be used standalone, with all other parts of the toolkit being uninitialized. The functionality is limited to loading and saving files, and reading and writing pixels (outside begin_paint only). All other calls are ignored.
This feature 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 the Prima::noX11 manpage for more.
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.
0,0xffffff
.
See also palette
.
ict::XXX
constants:
ict::None - no dithering ict::Ordered - 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::Ordered: [ 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 ]
::vScaling
property,
the pixel values are either scaled or truncated.
::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.
::sum
of pixel values, divided by number of pixels.
[0,0,0,255,255,255]
.
See also colormap
.
Prima::Drawable::pixel
.
end_paint()
.
::vScaling
and ::hScaling
properties,
the pixel values are either scaled or truncated.
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
.
::variance
.
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 im::SignedInt
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
::sum2
, divided by number of pixels
minus square of ::sum
of pixel values.
::hScaling
property,
the pixel values are either scaled or truncated.
::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.
::autoMasking
set to am::MaskColor
, COLOR
is used as a transparency value.
::autoMasking
set to am::MaskIndex
, INDEXth
color in teh current palette is used as a transparency value.
get_bpp()
method; monochrome
bitmaps always have bit depth of 1.
See the Prima::image-load manpage for details.
This method can be called without object instance.
::type & im::BPP
.
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 the Prima::image-load manpage for details.
NB! When loading from streams on win32, mind binmode
.
::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 ]
$image-> resample( $image-> rangeLo, $image-> rangeHi, 0, 255);
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 the Prima::image-load manpage for details.
NB! When saving to streams on win32, mind binmode
.
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 the Prima::image-load manpage for details, watch_load_progress in the Prima::ImageViewer manpage and
load in the Prima::ImageDialog manpage for suggested use.
load
option eventDelay
to limit the rate of DataReady
event.
::data
storage,
in the second - from ::mask
storage.
::data
and ::mask
property. DATA and MASK are expected to be images
of same dimension.
Dmitry Karasik, <dmitry@karasik.eu.org>.
Prima, the Prima::Drawable manpage, the Prima::image-load manpage, the Prima::codecs manpage.
PDL, the PDL::PrimaImage manpage, IPA
ImageMagick, the Prima::Image::Magick manpage
Prima::Image - Bitmap routines |