| AArch64 | |
| ppc64le | |
| s390x | |
| x86-64 |
- Update to version 2.6.1 * New functions. * Better MATLAB compatibility.
- Update to version 2.4.1 * Bugfix release - Drop ovsolete workaround_gcc_65843.patch
- Fix build with GCC5 * workaround_gcc_65843.patch
- Update to version 2.4.0
* The following functions are new:
bwareafilt imcast imregionalmin
bwpropfilt imclearborder otf2psf
edgetaper immse psf2otf
fftconvn imreconstruct psnr
imattributes imregionalmax subimage
* The implementation of normxcorr2 has been changed. The new method is
Matlab compatible and will return values in the range [-1 1].
* The image package is no longer dependent on the signal package.
* The disk shaped filter of fspecial has been changed for Matlab
compatibility. The elements on the border of the disk are now
weighted by how much of them is covered by the disk. Note that
this change is backwards incompatible.
* The following functions will display the output image as a figure
instead of printing to the command line, when there are no output
arguments:
grayslice im2bw
* For better compatibility with Matlab, the imrotate, imremap, and
imperspectivewarp functions now use 0 instead of NA for the default
extrapolation value.
* The regionprops function now supports the "Eccentricity",
"MajorAxisLength", "MinorAxisLength", "EquivDiameter" and "Extrema"
properties. The "Orientation" property has also been rewritten for
Matlab compatibility and may yield different results than previous
versions.
* The conndef function has new function signatures so that it covers all
common ways of defining an connectivity array.
* Creating disk shaped strel objects must now specify the N argument for
number of periodic lines used to approximate a disk. The value used
must be zero to obtain the same results as previous releases of the
image package. No other value is at the moment valid, and this is to
prevent future backwards incompatibility since Matlab default is 4.
Replace any `strel ("disk", radius)' with `strel ("disk", radius, 0)'
* The following functions have been completely rewritten and will perform
a lot of faster.
bwconncomp bwlabeln
Which indirectly will also cause the following to perform faster:
bwareaopen bwpropfilt regionprops
* Deprecated functions.
The following functions were deprecated in image 2.2.0 and have been
removed from image 2.4.0.
bwborder iptchecknargin readexif
impad iptcheckstrs
imrotate_Fourier uintlut
* Other functions that have been changed for smaller bugfixes, increased
Matlab compatibility, or performance:
grayslice im2single imsmooth
im2double im2uint8 label2rgb
im2int16 im2uint16
- Split from octave-forge package, version 2.2.2
- Update to version 2.14.0:
* The following functions are new: affine, affine2d, affine3d,
imapplymatrix, imfuse, imshowpair.
* New features "anti-aliasing" and "custum kernel" were added to
imresize. New method "box" was also added.
* Inaccuracies in imresize bilinear interpolation and in bicubic
interpolation were fixed, making imresize much more compatible
with Matlab.
* The function grayslice returns different results on input
images of class int16.
* Other functions that have been changed for smaller bugfixes,
increased Matlab compatibility, or performance: psf2otf,
imtransform.
- Update to version 2.12.0:
* The following functions are new:
imref2d integralImage integralImage3
imref3d
* Improved the cubic interpolation method in imremap. This should
also cause improved results in the functions imperspectivewarp,
imresize, and imrotate when using the cubic or bicubic methods.
* The function labelmatrix was changed for compatibility with the
upcoming Octave version 6.
* The function bwmorph now supports the "endpoints" operation.
* The function nonmax_supress was renamed nonmax_suppress to fix its
incorrect spelling. The incorrectly named function is kept for
backwards compatibility.
* The functions imremap, imperspectivewrap, and imrotate no longer
return the `valid' output argument.
* The imgradientxy function will now recognize the methods "central"
and "intermediate" for Matlab compatibility. Their previous names
"centraldifference" and "intermediatedifference" continue to work
for backwards compatibility.
* The stdfilt function will again handle all images of non floating
type (a regression introduced in the version 2.10.0).
- Update to version 2.10.0:
* The following functions are new: axes2pix, colorangle, houghpeaks
bwpack, deconvwnr, otsuthresh, bwunpack, houghlines
* The functions entropyfilt, rangefilt, and stdfilt were completely
incorrect and have now been fixed.
* In function regionprops, "Perimeter" was changed to be compatible
with Matlab, and "PerimterOld" was added.
* In function regionprops, properties "ConvexHull", "ConvexImage",
"ConvexArea", and "Solidity" were added.
* In function bwfill, the parameter connectivity was changed to be
Matlab compatible.
* grayslice now expects argument V to be a vector with the actual
values used to perform threshold.
* The function operation "thin" in bwmorph was changed to be Matlab
compatible.
* imcast now also converts images to logical type.
* wiener2 is now able to handle images with arbitrary number of
dimensions.
* Other functions that have been changed for smaller bugfixes,
increased Matlab compatibility, or performance: bwdist, entropy
hough_line, bwmorph, graythresh
Read /usr/share/octave/packages/image-2.10.0/packinfo/NEWS for
more details
- Update to version 2.8.1:
* Fix imcrop for non-square images, a regression introduced in
version 2.8.0.
- Update to version 2.8.0:
* The following functions are new: hough, imhmax, ntsc2rgb,
imextendedmax, imhmin, rgb2ntsc, imextendedmin, imimposemin,
viscircles, imfindcircles, imsharpen, wiener2.
* Octave version 4.0 is no longer supported. This makes Octave
version 4.2.0 the minimum requirement.
* The functions ntsc2rgb and rgb2ntsc have been removed in Octave
core version 4.4 and are now part of the image package. In a
similar manner, rgb2gray, which has always been part of the image
package, is part of Octave core version 4.4. The versions in
the image package will only be installed if they are missing
from Octave to avoid shadowing the core functions.
* imcrop will now adjust the region to crop to the image
boundaries when the region to crop goes beyond the image
boundaries.
* ycbcr2rgb, rgb2ycbcr, and rgb2gray will return an image of
class single if the input was class single (they were previously
returning a double for such cases).
* The function imdither has been removed. This function hadn't
work since Octave version 3.8.0 and would need to be
rewritten. The function dither from Octave core should be
used instead but hasn't been implemented yet. Instead of
rewriting imdither, the missing dither function should be
written. See bug #41768.
* graythresh will now clip values outside the [0 1] range in
images of floating point class. A side effect of this is that
it will no longer distinguish between an image of floating
point class that has only one one row or column, and an
histogram. Breaks backwards compatibility if you were
previously relying on something like 'graythresh (img(:),
...)' and img is of class single or double.
* Fix regionprops Perimeter when using a labelled image or a
bwconncomp structure, from connectivity different from 8.
* The functions ycbcr2rgb and rgb2ycbcr now support conversion
according to the BT.2020 standard.
* Other functions that have been changed for smaller bugfixes,
increased Matlab compatibility, or performance: bwperim,
imlincomb, imrotate, entropy, imreconstruct, normxcorr2,
graythresh, imregionalmax, phantom, imhist, imregionalmin,
rgb2gray.
- Update (Build)Requires: octave >= 4.2.0 now required.