A C++11 library for efficient hierarchical image segmentation.

Please cite the following papers accordingly if you use the code:

• T. Liu, C. Jones, M. Seyedhosseini, T. Tasdizen. A modular hierarchical approach to 3D electron microscopy image segmentation. Journal of Neuroscience Methods, 226, pp. 88--102, 2014.

• T. Liu, E. Jurrus, M. Seyedhosseini, T. Tasdizen. Watershed merge tree classification for electron microscopy image segmentation. ICPR 2012.

• T. Liu, M. Seyedhosseini, T. Tasdizen. Image segmentation using hierarchical merge tree. IEEE Transactions on Image Processing, 25, pp. 4596--4607, 2016.

• T. Liu, M. Zhang, M. Javanmardi, N. Ramesh, T. Tasdizen. SSHMT: Semi-supervised hierarchical merge tree for electron microscopy image segmentation. ECCV 2016.

Use a modern compiler with C++11 support, e.g., GCC-4.8 or higher and Apple LLVM 6.

Dependencies:

• InsightToolkit (ITK).
• Boost C++ libraries.
• Eigen.

Instructions:

• Use '-DCMAKE_CXX_FLAGS=-std=c++11' for the first time ITK CMake configuration.
• Turn on 'ITKReview' module for ITK.
• Enable C++11 for Boost libraries.

CMake configurations:

• Turn on 'GLIA_MT' to use OpenMP parallelization.
• Work on 3D/2D images with 'GLIA_3D' turned on/off.
• Turn on 'GLIA_BUILD_{HMT,SSHMT,LINK3D,GADGET,ML_RF}' modules accordingly.
• The random forest classifier used in our code is based on Abhishek Jaiantilal's R-to-MATLAB migration (https://github.com/ajaiantilal/randomforest-matlab) of random forest. To use the related functionalities, please turn on 'GLIA_BUILD_ML_RF' and 'GLIA_HMT_USE_RF', and set 'RF_SRC_DIR' as the path to 'RF_Class_C/src/' folder in their code.

• Ting Liu <[email protected]>