Superpixels through Iterative CLEarcutting (SICLE)

This is the implementation of the superpixel segmentation method Superpixels through Iterative CLEarcutting(SICLE) as proposed in:

• F.Belém, B.Perret, J.Cousty, S.Guimarães, A.Falcão. Efficient Multiscale Object-based Superpixel Framework. ArXiv preprint.
• F.Belém, I.Borlido, L.João, B.Perret, J.Cousty, S.Guimarães, A.Falcão. Fast and Effective Superpixel Segmentation Using Accurate Saliency Estimation. Discrete Geometry and Mathematical Morphology, pg. 261-273, 2022.
• F.Belém, I.Borlido, L.João, B.Perret, J.Cousty, S.Guimarães, A.Falcão. Novel Arc-Cost Functions and Seed Relevance Estimations for Compact and Accurate Superpixels. Journal of Mathematical Imaging and Vision, 2023.

This software includes nine programs:

• One for creating an object's minimum bounding box image from a ground-truth;
• One for creating an image whose superspels are colored by their mean color;
• One for assisting the visualization of the segmentation by overlaying the superspel borders;
• One for creating an image whose superspels are pseudo-colored;
• One for relabelling the superspels from a superspel segmentation;
• One for running a SICLE method;
• One for running the SICLE method in its optimized compact version;
• One for running the SICLE method in its optimized irregular version;
• One for assessing a superspel segmentation;

Please cite the aforementioned papers (preferably the last one) if you use any of this software in your own project.

Hardware, Setup and Requirements

The project was developed in C under a Linux-based operational system; therefore, it is NOT GUARANTEED to work properly in other systems (e.g. Windows and macOS). Moreover, the same applies for non-GCC compilers, such as Clang and MinGW.

All code within this project were developed, compiled and tested using the following programs:

• GNU GCC: version 7.5.0
• GNU Make: version 4.1

This code was implemented and evaluated in a computer with the following specifications:

• CPU: 64-bit Intel(R) Core(TM) i7-4790 @ 3.20 GHz
• Memory: 8GB RAM

The library has in-built support for handling PNM images. For enabling external library support, please refer to the README file within the externals folder.

Compiling

If your computer meets the aforementioned requirements, you may run the commands below for compiling the library and all the demonstration programs.

make lib
make demo

Or simply run one of the following commands for compiling both latter at once.

make
make all

For removing the files generated from compilation, one may run the following rule.

make clean

Running

After compiling, one may run any of the demonstration programs in demo folder, such as RunSICLE with the option --help for outputting the possible parameters for that program, as it is shown below:

./bin/RunSICLE --help

Then, the program will indicate a brief description of all parameters (both required and optional) for its proper execution.

License

All codes within this project are under the MIT License. See the LICENSE file for more details.

Acknowledgements

This work was financially supported by the following funding agencies:

• Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq)
• Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)
• Fundação de Amparo à Pesquisa do Estado de Minas Gerais (FAPEMIG)
• Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

Contact

If you have any questions or faced an unexpected behavior (e.g. bugs), please feel free to contact the authors through the following email addresses:

• Felipe C. Belém: felipe.belem@ic.unicamp.br or felipe.belem@esiee.fr
• Benjamin Perret: benjamin.perret@esiee.fr
• Jean Cousty: jean.cousty@esiee.fr
• Silvio Jamil F. Guimarães: sjamil@pucminas.br
• Alexandre X. Falcão: afalcao@ic.unicamp.br