21 March 2024

Changes DIPlib 3.4.2

Changes to DIPlib

Added "Oklab", "Oklch", "Y'PbPr" and "Y'CbCr" as color spaces.
Added dip::Distribution::ConstSample as an unmutable version of dip::Distribution::Sample.
Added functions dip::ImageReadPNG(), dip::ImageReadPNGInfo(), dip::ImageIsPNG() and dip::ImageWritePNG(). Previously, PNG files could only be read through DIPjavaio with Bio-Formats.
The functions dip::ImageRead() and dip::ImageWrite() now recognize PNG files.
Added overloads for dip::ImageReadPNG(), dip::ImageReadPNGInfo(), dip::ImageWritePNG(), dip::ImageReadJPEG(), dip::ImageReadJPEGInfo(), and dip::ImageWriteJPEG() that read from and write to a memory buffer rather than a file.
Added function dip::FileAppendExtension().

The ‘diverging’ color map in dip::ApplyColorMap() switched from CET-D08 to CET-D07 (i.e. using yellow instead of red for positive values).
dip::Distribution::Iterator is now a template class that takes either dip::Distribution::Sample or dip::Distribution::ConstSample as argument. dip::Distribution::MutableIterator and dip::Distribution::ConstIterator are two template specializations, the first one replaces the use of dip::Distribution::Iterator.
dip::FileAddExtension() has been deprecated, the functionality is not needed in DIPlib.

XYZ to Yxy conversion and its inverse were reversed, and Yxy to gray conversion was incorrect. This means that the Yxy color space was all wrong and inconsistent.
NPY file header written was correct according to the specification, but didn’t exactly match what was written by NumPy itself. NumPy under Python 3.12 errored out reading these files. Now the header exactly matches NumPy’s.
dip::SeededWatershed() and dip::CompactWatershed() discarded the seeds input image when it was the same as the out image (when the caller intended for the algorithm to work in-place).
Member functions of dip::Distribution return ConstSample or ConstIterator when the distribution object is const. See issue #135.
DIPlib 3.4.1 changed the behavior of generating a view of an image using a mask when the mask was empty (i.e. selected no pixels). This fixed one issue, but created many more: img.At(mask)=0 stopped working when the mask happened to be empty, which is unexpected and very inconvenient (as this syntax is quite common). Now an empty view can be generated again, but when cast to an image, a raw image is generated.
Equality comparison of dip::PixelSize objects was documented to use a tolerance, but it wasn’t using one.
dip::ImageReadTIFF(), dip::ImageReadJPEG() and dip::ImageReadNPY() would replace the extension of the given file name instead of simply appending an appropriate extension, as described in the documentation.

Updated included zlib to version 1.3.1. This version no longer uses K&R function declarations, which generated lots of warnings in newer compilers.

New function apply_colormap calls dip::ApplyColorMap().
The color maps used in the dipshow tool are now directly taken from DIPlib. This changes the ‘periodic’ and ‘diverging’ color maps.
The dipshow ‘Mappings’ menu has a new option ‘Sequential’, which applies the "linear" color map from dip::ApplyColorMap(). This also affects dipmapping, which has a new option 'sequential'.

(See also changes to DIPlib.)

joinchannels produced an error if the first input image was a MATLAB array rather than a dip_image object.
Dragging viewslice windows between screens no longer causes MATLAB to crash.

(See also bugfixes to DIPlib.)

Added functions dip.ImageReadPNG(), dip.ImageReadPNGInfo(), dip.ImageIsPNG() and dip.ImageWritePNG().
Added functions for reading and writing PNG and JPEG as a bytes object.

dip.Show() now uses the DIPlib color maps by default, taken from dip.ApplyColorMap(). The names recognized by that function can be used for the colormap parameter. Other names are still passed to Matplotlib.
All dip.ImageRead...() functions now have an overload with named out argument to read an image into a pre-allocated memory buffer.

(See also changes to DIPlib.)

None, but see bugfixes to DIPlib.

The “RGB” display will convert the colors in the image to sRGB, rather than linear RGB, to match the behavior in other parts of the DIPlib project.

SliceViewer properties given as DimensionArrays are now properly converted to Python buffers again.
Maintaining a SliceViewer handle after a window has been closed no longer causes a crash.

None