This is the code I used on my thesis "Efficient stereo matching and obstacle detection using edges in images from a moving vehicle" where image edges are used to represent the world in a compact manner.
The code has been tested on an Ubuntu 16.04 system installed on an Intel i72675QM CPU. The project is built using cmake and requires a working version of OpenCV 2, pkg-config, Eigen3, boost and libpng.
Install the dependences
sudo apt-get install cmake pkg-config libeigen3-dev libpng++-dev libboost-all-dev
Follow the guide for installing OpenCV 2 from source.
The folder testing contains a sample setup to run get the stixels from a stereo pair. You can take this as reference for your implementations. Currently I am working on a version without any dependences but it is not available yet.