Library for Efficient Large-Scale Stereo Matching (ELAS) with OpenCV support

This is a port of the libelas (LIBrary for Efficient LArge-scale Stereo matching) C++ library for stereo matching to work with OpenCV and support the meson build system additionally to CMake. The library implements a stereo matching method proposed in the following publication:

@INPROCEEDINGS{Geiger10,
 author = {Andreas Geiger and Martin Roser and Raquel Urtasun},
 title = {Efficient Large-Scale Stereo Matching},
 booktitle = {Asian Conference on Computer Vision},
 year = {2010},
 month = {November},
 address = {Queenstown, New Zealand}
}

If you use this library, please cite the above paper.

Differences

• The CMakeLists.txt was adapted to include the OpenCV dependency and to build a library target. Further, the example file is excluded from the default build targets and needs to be build explicitly. Additionally, compile flags for Visual Studio were added.
• Meson build support was added (meson.build and meson_options.txt file).
• The PROFILE flag was added to both build systems as command line argument so that it can be switched on and off without changing anything in the code.
• Include files were moved to own include directory.
• Two functions to convert images from and to OpenCV matrices (imageFromCvMat and cvMatFromImage) were implemented in image.h to add support for OpenCV.
• The elas namespace was added to elas.h, elas.cpp and image.h.
• Compile warnings in all files were handled (see comments in header of files) and Visual Studio specific defines for fixed-width data types were removed since they are not required anymore.
• Matlab support was removed.

Building

Prerequisites

• C++ compiler
• meson or CMake
• ninja or make
• OpenCV C++ library

Build Instructions

• Move to project root directory

Meson

• Run meson to initialize build directory: meson build
• Build library and example with ninja: ninja -C build

CMake

• Create build directory and enter it: mkdir build && cd build
• Run cmake: cmake .. && cd ../
• Build library: make -C build
• Build example: make -C build example

Compile Options

• With the elas_profiler build flag, the runtime measurement of the elas library can be activated.

Run example

• Run './build/example demo' => computes disparity maps for images from the img directory.