A quality metric for lossy image and video compression.
This is a port of the VapourSynth plugin butteraugli.
libjxl's Butteraugli is used.
-
AviSynth+ 3.6 or later
-
Microsoft VisualC++ Redistributable Package 2022 (can be downloaded from here)
Butteraugli (clip reference, clip distorted, bool "distmap", float "intensity_target", bool "linput")
-
reference, distorted
Clips that are used for estimating the psychovisual similarity.
They must be in RGB 8/16/32-bit planar format and must have same dimensions. -
distmap
Whether to return heatmap instead of distorted clip
Default: False. -
intensity_target
Viewing conditions screen nits.
Default: 80.0. -
linput
True: The input clips must have linear transfer functions.
False: The input clips are assumed in sRGB color space and internally converted to linear transfer.
Default: False.
The psychovisual similarity of the clips will be stored as frame property '_FrameButteraugli' in the output clip. Larger values indicate to bigger difference.
Requirements:
- Git
- GCC C++17 compiler
- CMake >= 3.16
- Ninja
git clone --recurse-submodules https://github.com/Asd-g/AviSynthPlus-Butteraugli
cd AviSynthPlus-Butteraugli\libjxl
cmake -B build -DCMAKE_INSTALL_PREFIX=jxl_install -DCMAKE_BUILD_TYPE=Release -DBUILD_TESTING=OFF -DJPEGXL_ENABLE_FUZZERS=OFF -DJPEGXL_ENABLE_TOOLS=OFF -DJPEGXL_ENABLE_MANPAGES=OFF -DJPEGXL_ENABLE_BENCHMARK=OFF -DJPEGXL_ENABLE_EXAMPLES=OFF -DJPEGXL_ENABLE_JNI=OFF -DJPEGXL_ENABLE_OPENEXR=OFF -DJPEGXL_ENABLE_TCMALLOC=OFF -DBUILD_SHARED_LIBS=OFF -G Ninja
cmake --build build
cmake --install build
cd ..
cmake -B build -DCMAKE_BUILD_TYPE=Release -G Ninja
cmake --build build