IAM

OPENSOURCEFAN STUDYING
STUDYINGCOMPUTERSCIENCEANDMATH COMPUTERSCIENCE

Check out the latest superpixel benchmark — Superpixel Benchmark (2016) — and let me know your opinion! @david_stutz
15thJULY2014

READING

R. Achanta, A. Shaji, K. Smith, A. Lucchi, P. Fua, S. Süsstrunk. SLIC Superpixels Compared to State-Of-The-Art Superpixel Methods. Transactions on Pattern Analysis and Machine Intelligence, pages 2274 – 2282, 2012.

Achanta et al. proposed their superpixel algorithm "Simple Linear Iterative Clustering", short SLIC, in [1]. The above paper compares SLIC to several superpixel algorithms with respect to Boundary Recall, Undersegmentation Error and runtime: Normalized Cuts [2], the approach proposed by Felzenswalb and Huttenlocher [3], Quick Shift [4], Watersheds [5] and Turbopixels [6].

Superpixel segmentations generated by the original implementation of SLIC can be found in figure 1. Additionally, the VLFeat Library [7] provides an implementation of SLIC, see my article on running VLFeat's implementation of SLIC using C++ and CMake.

Update. Thorough evaluation of both implementations can be found in my bachelor thesis: Bachelor Thesis “Superpixel Segmentation Using Depth Information”.

bsd_1_slic_compactness_1 bsd_1_slic_compactness_1 bsd_1_slic_compactness_1 bsd_1_slic_compactness_1 bsd_1_slic_compactness_1
bsd_1_slic_compactness_1 bsd_1_slic_compactness_1 bsd_1_slic_compactness_1 bsd_1_slic_compactness_1 bsd_1_slic_compactness_1
bsd_1_slic_compactness_1 bsd_1_slic_compactness_1 bsd_1_slic_compactness_1 bsd_1_slic_compactness_1 bsd_1_slic_compactness_1

Figure 1: Superpixel segmentations with roughly $600$ superpixels generated by the original implementation of SLIC which allows to adjust the compactness of the superpixels. The images are taken from the validation set of the Berkeley Segmentation Dataset [8]. From top to bottom: compactness set to $1$; compactness set to $10$; compactness set to $40$.
  • [1] R. Achanta, A. Shaji, K. Smith, A. Lucchi, P. Fua, S. Susstrunk. SLIC Superpixels. Technical report, EPFL, Lausanne, 2010.
  • [2] X. Ren, J. Malik. Learning a classification model for segmentation. Proceedings of the International Conference on Computer Vision, pages 10–17, 2003.
  • [3] P. F. Felzenswalb, D. P. Huttenlocher. Efficient graph-based image segmentation. International Journal of Computer Vision, volume 59, number 2, 2004.
  • [4] A. Vedaldi, S. Soatto. Quick shift and kernel methods for mode seeking. Proceedings of the European Conference on Computer Vision, pages 705–718, 2008.
  • [5] L. Vincent, P. Soille. Watersheds in digital spaces: An efficient algorithm based on immersion simulations, Transactions on Pattern Analysis and Machine Intelligence, volume 13, number 6, pages 583-598, 1991.
  • [6] A. Levinshtein, A. Stere, K. N. Kutulakos, D. J. Fleet, S. J. Dickinson, K. Siddiqi. TurboPixels: Fast superpixels using geometric flows. Transactions on Pattern Analysis and Machine Intelligence, volume 31, number 12, pages 2290–2297, 2009.
  • [7] A. Vedaldi, B. Fulkerson. VLFeat: An open and portable library of computer vision algorithms. http://www.vlfeat.org/, 2008.
  • [8] P. Arbeláez, M. Maire, C. Fowlkes, J. Malik. Contour detection and hierarchical image segmentation. Transactions on Pattern Analysis and Machine Intelligence, volume 33, number 5, pages 898–916, 2011.

What is your opinion on the summarized work? Or do you know related work that is of interest? Let me know your thoughts in the comments below or get in touch with me: