A sensitivity analysis of initial conditions for the Mumford-Shah based level set method of image segmentation (student paper)

ABSTRACT

The level set method provides a means of segmenting images. Fundamentally, the level set method for image segmentation is a search algorithm that determines where an evolving curve's boundary pixels - who are meant to encompass an image segment's perimeter - should be placed according to some criteria. A method has been devised that utilizes the Mumford-Shah functional as a means of establishing that criteria. It has been shown that this method for image segmentation has limitations that, while mentioned in the original research, were not quantified with examples. We present evidence of these limitations, discuss how they occurred and describe our preliminary attempts at overcoming them. We conclude by offering a direction where this research might lead. This work is part of an ongoing research project.

REFERENCES

• 1: Chan, T. and Vese, L. Active Contours without Edges. IEEE Trans. Of Image Processing, 10 (2):266--277, Feb. 2001
• 2: Kass, M., Witkin A., and Terzopoulos, D., "Snakes: Active contour models," Int. J. Comput. Vis., Vol. 1, pp. 321--331, 1988.
• 3: Trish Keaton , Jeffrey Brokish, A Level Set Method for the Extraction of Roads from Multispectral Imagery, Proceedings of the 31st Applied Image Pattern Recognition Workshop on From Color to Hyperspectral: Advancements in Spectral Imagery Exploitation, p.141-150, October 16-18, 2002
• 4: David R. Martin , Charless C. Fowlkes , Jitendra Malik, Learning to Detect Natural Image Boundaries Using Local Brightness, Color, and Texture Cues, IEEE Transactions on Pattern Analysis and Machine Intelligence, v.26 n.5, p.530-549, May 2004 [doi>10.1109/TPAMI.2004.1273918]
• 5: Mumford, D. and Shah, J. Optimal approximation by piecewise smooth functions and associated variational problems. Comm. On Pure and Appl. Math., XLII: 577--685, 1989.
• 6: Stanley Osher , James A. Sethian, Fronts propagating with curvature-dependent speed: algorithms based on Hamilton-Jacobi formulations, Journal of Computational Physics, v.79 n.1, p.12-49, Nov. 1988 [doi>10.1016/0021-9991(88)90002-2]
• 7: Yongsheng Pan , J. Douglas Birdwell , Seddik M. Djouadi, Bottom-Up Hierarchical Image Segmentation Using Region Competition and the Mumford-Shah Functional, Proceedings of the 18th International Conference on Pattern Recognition, p.117-121, August 20-24, 2006 [doi>10.1109/ICPR.2006.339]
• 8: Yongsheng Pan , J. Douglas Birdwell , Seddik M. Djouadi, Efficient Bottom-Up Image Segmentation Using Region Competition and the Mumford-Shah Model for Color and Textured Images, Proceedings of the Eighth IEEE International Symposium on Multimedia, p.376-390, December 11-13, 2006 [doi>10.1109/ISM.2006.64]
• 9: Pan, Y., Birdwell, J. D., Djouadi S. M., "Efficient Implementation of the Chan-Vese Models Without Solving PDEs," Proceedings of the International Workshop On Multimedia Signal Processing 2006, Victoria, BC, Canada, Oct. 03--06, 2006.
• 10: Seedahmed, G. and Martucci, L., "Automated Image Registration Using Geometrically Invariant Parameter Space Clustering (GIPSC)," Archives of the International Society of Photogrammetry and Remote Sensing, 2002.
• 11: Sethian, J. A. "An Analysis of Flame Propagation", PhD Dissertation, Dept. of Mathematics, Univ. of California, Berkeley, 1982.
• 12: Sethian, J. A. Tracking Interfaces with Level Sets, American Scientist, May-June, 1997.
• 13: Wasilewski, M. Active Contours using Level Sets for Medical Image Segmentation, http://www.postulate.org/segmentation/segmentation.pdf
• 14: http://en.wikipedia.org/wiki/Hamilton-Jacobi_equations
• 15: http://vuiis.vanderbilt.edu/~licm/research/LBF.html

INDEX TERMS

Primary Classification:
  I. Computing Methodologies
  I.m MISCELLANEOUS

General Terms:
Algorithms, Performance

Keywords:
image segmentation, measures, performance