SphSPS: Spherical Shortest Path-based Superpixels

Overview

Implementation of paper: PDF

@InProceedings{giraud2020sphsps,
    author   = {Rémi Giraud and Rodrigo Borba Pinheiro and Yannick Berthoumieu},
    title    = {Generalization of the Shortest Path Approach for Superpixel Segmentation of Omnidirectional Images},
    booktitle  = {Pattern Recognition (PR)},
    YEAR   = {2023},
}

• Segmentation of 360° spherical images into regular superpixels:

• Generalized Global Regularity Metric (G-GR) for evaluation of spatial superpixel regularity in the spherical space based on:

@InProceedings{giraud2017gr,
    author   = {Rémi Giraud and Vinh-Thong Ta and Nicolas Papadakis},
    title    = {Evaluation Framework of Superpixel Methods with a Global Regularity Measure,
    booktitle  = {Journal of Electronic Imaging},
    YEAR   = {2017},
}

Requirements

• Linux

• For C++ version: CImg library (unique .h already included)

• A contour prior map can be provided to our method (for C++: an image with 255 for highest contour intensity)
  The contour detection method used in this work is available here
  Other contour detection methods can be found here

Execution

MATLAB / C-Mex

run main.m    %call SphSPS_mex

C++

• Compilation:

make

• Execution prototype:

./SphSPS -i img_name [-k superpixel_nbr(1000)] [-m compactness(0.12)]  [-outm output_map_name(res/labelmap.png)] [-outb output_border_name(res/borders.png)]  [-c contour(NULL)]

• Execution with contour map: (make test)

./SphSPS -i ./data/test_img1.jpg -k 1500 -m 0.12 -outm test_img1_map.png -outb test_img1_border.png -c ./data/test_img1_contour.png

• Execution on an image list: (make test_list)

./scripts/test_list.sh ./data/list_file.txt ./data/ 1000 0.12

Data

The dataset of 75 manually annotated spherical images of size 512x1024 pixels capturing panoramas is available here
These images are extracted from the SUN360 dataset

License

(C) Rémi Giraud, 2020
[email protected]
https://remi-giraud.enseirb-matmeca.fr
ENSEIRB-MATMECA (Bordeaux INP), Laboratory IMS

This code is free to use, share and modify for any non-commercial purposes. Any commercial use is strictly prohibited without the authors' consent, also including authors from (chen2017) since SphSPS uses some part of their code:

@InProceedings{chen2017,
    author   = {Jiansheng Chen and Zhengqin Li and Bo Huang},
    title    = {Linear Spectral Clustering Superpixel},
    booktitle  = {IEEE Transactions on Image Processing},
    YEAR   = {2017},
}