Changelog

List of changes in each version

v1.14.0

• Fix mono bmp reader and writer

v1.13.1

• Fix dynamic loader broken when included without being enabled

v1.13

• Add AVX2 color correction for RGB8 and BGR8 and optimized RGBa8 and BGRa8
• Fix some return codes having the wrong prefix when converted to a string using StringFromPEAK_IPL_RETURN_CODE

v1.12.1

• Fix a crash when the factors are changed while a binning instance currently bins an image
• Fix ARM NEON debayering in fast mode

v1.12

• Add the Mono16 pixelformat
  • It can be converted from and into
  • Can be saved and read from/to png, bmp and tiff files
• Add IsFloat Function to check if a pixelformat uses the floating point data type
• Add support for writing and reading a limited number of TIFF (Tagged Image File Format) files
  • No compression and compressions according to TIFF Baseline are supported
  • Supported color spaces: RGB and black as zero
  • Support for 8 and 16-Bit per sample (10 and 12-Bit are supported by the Tag TIFFTAG_BITSPERSAMPLE)
  • Image construction from stripes
• Python Bindings:
  • The bindings now embed their direct runtime dependencies. With this change the python package can be used without requiring any setup to be installed.

v1.11

• Fix debayer artifacts for fast conversion with image widths not divisible by 8 on some embedded systems and when converting into mono formats
• Change the python bindings to use the Python Limited API
  • The minimum required python version is 3.11. They can be installed on any later 3.x version
• Fix an error that prevented the use of the dynamic loader

v1.10

• Add support for pixelformat PixelFormatName::YUV422_8_UYVY
• Fix debayer artifacts for fast conversion with image widths not divisable by 8 (10 and 12 bit AVX2)

v1.9.3

• Add digital signature to dotnet assemblies
• Fix debayer artifacts for fast conversion with image widths not divisable by 8 (8 bit AVX2)
• Fix SharpnessAlgorithm::HistogramVariance returning NaN for black images

v1.9.2

• Add static functions to peak::ipl::Library to configure multithreading and simd, and get the library version.

v1.9.1

• Fix [ImageConverter::SupportedOutputPixelFormatNames] to return only the supported pixel formats for the set conversion mode.
• Fix reading beyond the source buffer for specific image sizes when using one of the SIMD unpacking algorithms

v1.9

• Add support for pixelformat PixelFormatName::YUV420_8_YY_UV_SemiplanarIDS and PixelFormatName::YUV420_8_YY_VU_SemiplanarIDS
• Fix handling of 8-bit PNG files.

v1.8

• Add method PixelFormat::BayerPattern() to receive the corresponding BayerPattern.
• Add RGB/BGR(A) to mono converter
• Add reference counting to classes, so they can be shallow copied
• Improve the performance of the histogram
• Improve performance of mirror and rotate inplace functions
• Fix compilation error with dynamic loader.

v1.7

• Add a EdgeEnhancement implementation
• Changed the dotnet bindings to use .NET Standard 2.0.
  • See https://learn.microsoft.com/de-de/dotnet/standard/net-standard?tabs=net-standard-2-0#select-net-standard-version
  • Note the changed .NET Version requirement:
    • .NET Framework 4.6.1 and later
    • .NET Core 2.0 and later (including .NET 5 and later)
• Adjust move constructor and move assignment to not call the internal destruct function
• Add new methods [HotpixelCorrection::CorrectInPlace] and [HotpixelCorrection::CorrectAdaptiveInPlace] for in place hotpixel correction
• Add new methods [Image::Scale], [Image::ScaleInto] and [Image::Crop] for image processing
• Add multithreading support for detecting hotpixel within a given image.
• Add SSE2/AVX2 implementation for a Bayer10p/Mono10p to Bayer10/Mono10 decoder and Bayer12p/Mono12p to Bayer12/Mono12 decoder
• Optimized 8/10/12 bit BayerRG and BayerGR debayering. Added optimized 8/10/12 bit BayerGB and BayerBG debayering
• Optimization of image rotation and image mirroring
• Changed the python bindings to release the GIL for long-running functions
• Add NEON implementation for debayering and color correction
• Fix MSVC 2015 unable to compile some of the provided header implementations
• Fix unnecessary buffer copy in [ImageConverter::Convert] when passing an output buffer.
• Fix some functions of the binding not working correctly (notably Sharpness, VideoWriter and Histogram)
• Fix Histogram not working correctly
• Breaking changes
  • Renamed the member functions of the Range class

v1.6

• Add a signed .NET binding library for strong-named assemblies support
• Add saturation feature to ColorCorrector
• Add the Decimation algorithm for mono, bayer and some color images
• Add new operators for comparison to the Image class
• Add a Sharpness implementation which can be used to measure the sharpness of images
• Add multithreading support to the library and use it for some functions e.g. Sharpness
• Add method Image::Timestamp to receive the corresponding timestamp.
• Remove the RGB10p32 format from Binning as it was not actually supported
• Improve performance of 10/12 bit BayerRG and BayerGR debayering (x86 systems with SSE/AVX)
• Improve performance of 10/12 bit color correction (x86 systems with SSE/AVX)
• Fix HotpixelCorrection functions crashing if image dimension is smaller than 3 pixels for bayer images or smaller than 2 pixels for mono images
  • HotpixelCorrection functions will now throw an OutOfRangeException if the image dimensions are too small to process the image properly
• Fix images converted with Image::ConvertTo were actually converted inplace, when a destination image was given
• Fix images converted with Image::ConvertTo were not set to the correct alpha value, when channels were expanded.
• Fix wrong file size in the file header when writing bmp files
• Breaking changes
  • Specified the UTF-8 encoding for ImageWriter amd ImageReader

v1.5

• Add a VideoWriter implementation which can be used to save images in an avi container encoded as mjpeg

v1.4

• Add new function ImageConverter::PreAllocateConversion which can be used to pre-allocate the internal buffer pools of the converter in order to save time on the first conversion
• Add new function Binning::ExpectedOutputSize which can be used to calculate the output size after binning of a given input size and pixelformat
• New CMake packages for both windows and linux, which now take care of adding their respective include directories
• Add new CMake function ids_peak_ipl_deploy to take care of copying dependent files(i.e. DLL's) into your targets output directory
• Add corresponding function ids_peak_ipl_dotnet_deploy when using dotnet
• Add the Binning algorithm for mono, bayer, some color and RGB10p32 images
• Fix HotpixelCorrection algorithms not working on the borders of an image
• Fix ImageWriter saving 10 Bit formats, although not supported, resulting in a broken bmp file with wrong colors
• Fix an issue with user defined hotpixels
• Fix images converted with Image::ConvertTo not freed when leaving the scope
• Fix images cloned with Image::Clone not freed when leaving the scope
• Fix images converted with Image::ConvertTo taking up 5 times more memory than necessary
• Fix images cloned with Image::Clone taking up 5 times more memory than necessary
• Breaking changes
  • Due to previously mentioned fixes, the calls to Image::ConvertTo and Image::Clone are now allocating memory on every use, causing additional overhead and will take more time. Users are advised to construct an ImageConverter which keeps internal memory pools.

v1.3.2

• Fix ColorCorrector crashing application when used on proccessors, which do not support AVX instructions.
• Add the Gain algorithm for mono and bayer images
• Add method ImageConverter::SupportedOutputPixelFormatNames to .NET binding
• Add BufferTooSmallException for Image::ConvertTo and ImageConverter::Convert when providing an output buffer that is too small

v1.3.1

• Fix issues regarding peak_dynamic_loader.h
• Fix problem with ImageTransformer::RotateInPlace throwing exception when passing raw image buffers
• Add color correction support for pixel formats RGB10p32, RGB10 and RGB12
• Add gamma support for RGB10p32
• Fix breaking change: Add backwards compatibility for stdcall symbols

v1.3

• Add ConversionMode::Classic
• Add the GammaCorrector algorithm
• Fix ImageReader::Read() for invalid pixelformat not throwing an exception
• Fix problem with reading a raw image interpreted as a different possible pixelformat
• Add option for dynamic loading of shared library
• Add fixed ordinal numbers in Windows DLL, for future ABI compatibility
• Breaking changes
  • Change calling convention of the C interface functions from stdcall to cdecl.

v1.2

• Add Python binding
• Add methods for image rotation in 90° steps
  • ImageTransformer::Rotate()
  • ImageTransformer::RotateInPlace()
• Improve performance of BayerGR debayering
• Rework internal buffer management to more efficiently free unused buffers
• Fix possible memory leaks in PNG and JPEG related functions
• Fix PNG IO functions for 10-bit and 12-bit pixel formats
• Improve stability and error safety
• Improve Doxygen documentation
• Breaking changes
  • Move the .NET binding libraries to the newly introduced installation folder for bindings
  • Make PEAK_IPL_Library_GetVersionSubminor return the correct value
  • Make the non-default Image constructors reject creating images with a width or height of 0
  • Make Image::Clone() const
  • Make all functions which do not support packed pixel formats throw ImageFormatNotSupportedException
  • Throw exception on BGRa <-> RGBa conversion

v1.1

• Add .NET binding
• Add ColorCorrector algorithm
• Major performance improvement for various algorithms
• Breaking changes
  • Rename mirroring methods to reduce misunderstandings
    • MirrorVertical() to ImageTransformer::MirrorLeftRight()
    • MirrorHorizontal() to ImageTransformer::MirrorUpDown()
  • Rename PEAK_IPL_Library_GetVersionPatch() to PEAK_IPL_Library_GetVersionSubminor()
  • Rework the C interface in respect to consistency
• Various Bugfixes

v1.0

• Initial release