Librir

Librir is a C/C++/Python library dedicated to the manipulation of infrared videos. It is mainly used in the CEA/IRFM/WEST tokamak to archive and read back infrared videos acquired by the infrared diagnostic, and as a building block for Cognitive Vision/Machine Learning applications.

Librir is divided into 4 small libraries:

• tools: provides miscellaneous functions for file/folder management, datetime/string manipulation, etc...
• geometry: provides various functions to manipulate 2D polygons.
• signal_processing: provides various signal/image processing algorithms.
• video_io: record and read back compressed IR video files (lossy or lossless).
• registration (Python only): motion correction within infrared videos.

Librir can be used from C, C++ or Python, and is known to work on Windows (compiled with MingW or Visual Studio) and Linux (compiled with gcc).

While developped in C++14, Librir is mostly used through its Python interface. The Python wrapper is built using the low level C interface (using the C++ components) and the Python ctypes module.

Compiling Librir

Librir depends on the following C libraries:

• zstd: generic compression library.
• blosc: compression of numerical signals.
• libjpeg: lossy image compression.
• charls: lossy/lossless jpegls image compression.
• ffmpeg: video compression/decompression.

Librir directly embbeds the source code of these libraries which are compiled within the Librir compilation process. Only the ffmpeg library needs to be compiled separately. Note that Librir only supports versions 4.3 and 4.4 of ffmpeg, which must be compiled with at least libx264 and kvazaar support.

As configuring/compiling ffmpeg is a heavy error prone task, its compilation is automated by Librir compilation steps:

• For msvc builds, Librir ships a precompiled version of ffmpeg with all required dependencies (64 bits only).
• For gcc builds, the build process will automatically download ffmpeg, libx264 and kvazaar and compile them.

CMake is required to compile Librir. Below example shows how to compile Librir on Unix based environments:

git clone https://github.com/IRFM/librir.git
mkdir build
cd build
cmake ..
make
make install

This will install Librir locally in build/install. The install folder also contains the librir directory which is the python wrapper of the library. You can directly add the install folder to your Python library paths in order to use Librir from Python, or install it using the provided setup.py.

To perform a global installation, you must set the cmake option LOCAL_INSTALL to OFF. To ease the compilation process, the scripts build.sh (Unix environments and Msys2) and build.bat (msvc build) are provided. Type build --help for more information.

Using Librir from Python or C++

Installing Librir will produce the librir folder within the installation directory. This module is directly importable from Python as long as you add the installation path to your sys.path. You can also install it within your Python environment using the provided setup.py To use Librir from C/C++ code, a .pc file is available in the installation folder and can be consumed by pkg-config. For cmake users, Librir provides configuration files for the find_package function. Example:

cmake_minimum_required(VERSION 3.16)

project(test)

list(APPEND CMAKE_PREFIX_PATH "librir_installation_path/lib/cmake/librir")

find_package(librir REQUIRED)

include_directories("${LIBRIR_INCLUDE_DIR}")
link_directories("${LIBRIR_LIB_DIRS}")
add_executable(test test.cpp)
target_link_libraries(test ${LIBRIR_LIBRARIES} )

Using Librir in Labview

A Labview wrapper is also available. So far, it only implements limited functionalities, such as:

• converting PCR file to h264 lossless
• saving as .h264 files
• handling .h264 files

Authors

Librir is developed and maintained by:

• Victor MONCADA (victor.moncada@cea.fr)
• Leo DUBUS (leo.dubus@cea.fr)
• Erwan GRELIER (erwan.grelier@cea.fr)
• Christian STARON (christian.staron@cea.fr)