GR: Global Regularity metric for superpixel evaluation


Back to research page

Highlights


  • New Global Regularity (GR) metric to evaluate the regularity of superpixel decompositions
  • Both shape regularity and consistency are considered in GR
  • The GR metric addresses the non-robustness of state-of-the-art ones
  • Robustness to scale and noise, with higher correlation to performances of superpixel methods

Method


Shape regularity
  • New Shape Regularity Criteria (SRC): evaluation of convexity, balanced repartition and contour smoothness
  • Comparison to the Circularity (C) metric [3], that considers circular shapes and is too sensitive to contour smoothness:

More relevant regularity evaluation, and differentiation of shape groups
(higher value means higher regularity)




Robustness to scale of the proposed SRC metric



Shape consistency
  • New Smooth Matching Factor (SMF): Comparison of the superpixels to the average shape


  • Comparison to the Mismatch Factor (MF) [4], that is non-robust to large shape outliers:

More relevant evaluation of shape consistency with SMF (higher value means higher consistency)


Global Regularity
  • Global Regularity (GR): Combination of SRC and SMF to evaluate both shape regularity and consistency in one metric
  • See [1] for extensive evaluation of GR on superpixel performances
  • New evaluation of superpixel methods according to regularity (robustness to regularity parameter setting)






  • Matlab implementation of the global regularity (GR) measure

    Download here




Main (to cite)

  1. R. Giraud, V.-T. Ta and N. Papadakis: Evaluation Framework of Superpixel Methods with a Global Regularity Measure
    Journal of Electronic Imaging (JEI), Special issue on superpixels for image processing and computer vision, 2017   https://hal.archives-ouvertes.fr/hal-01519635/file/Giraud_JEI_GR.pdf   BibTex
Other

  1. R. Giraud, V.-T. Ta and N. Papadakis: Robust Shape Regularity Criteria for Superpixel Evaluation
    IEEE International Conference on Image Processing (ICIP), 2017   https://hal.archives-ouvertes.fr/hal-01510062/file/Giraud_SRC_ICIP17.pdf   BibTex  


  1. A. Schick, M. Fischer, and R. Stiefelhagen: Measuring and evaluating the compactness of superpixels
    International Conference on Pattern Recognition (ICPR), 2012
  2. V. Machairas, M. Faessel, D. Cardenas-Peña, T. Chabardes, T. Walter, and E. Decencière: Waterpixels
    IEEE Trans. on Image Processing (TIP), 2015
  3. M. Y. Liu, O. Tuzel, S. Ramalingam and R. Chellappa: Entropy rate superpixel segmentation
    IEEE Int. Conf. on Computer Vision and Pattern Recognition (CVPR), 2011
  4. R. Achanta, A. Shaji, K. Smith, A. Lucchi, et al.: SLIC superpixels compared to state-of-the-art superpixel methods
    IEEE Trans. on Pattern Analysis and Machine Intelligence (PAMI), 2012
  5. P. Buyssens, I. Gardin, S. Ruan and A. Elmoataz: Eikonal-based region growing for efficient clustering
    Image and Vision Computing, 2014
  6. J. Yao, M. Boben, S. Fidler and R. Urtasun: Real-time coarse-to-fine topologically preserving segmentation
    IEEE Int. Conf. on Computer Vision and Pattern Recognition (CVPR), 2015
  7. J. Chen, Z. Li and B. Huang: Linear spectral clustering superpixel
    IEEE Trans. on Image Processing, 2017
  8. R. Giraud, V.-T. Ta and N. Papadakis: Robust Superpixels using Color and Contour Features along Linear Path
        Computer Vision and Image Understanding, 2018 (accepted)

    See project page



Back to top