Hi, I confirm the bug to be really in qhull. I substituted the computation of the convex hull in scipy (qhull) with another library (vigranumpy) and thins fix the problem. If useful I might consider working on a cython implementation of the 2D convex hull algorithm and flash a pull request. Best Luca Il giorno martedì 21 agosto 2012 12:40:39 UTC+2, Luca Fiaschi ha scritto:

Hi all, I get a strange bug trying to uses regionprops on some images. Computing the hull of some objects result a internal error in the qhull library: Below you can find the error and and the set of points which triggers it:

Does anybody know how to fix it?

[[ 5. 12.5] [ 6. 6.5] [ 7. 4.5] [ 8. 4.5] [ 9. 4.5] [ 10. 5.5] [ 7. 4.5] [ 7. 5.5] [ 6. 6.5] [ 6. 7.5] [ 6. 8.5] [ 6. 9.5] [ 6. 10.5] [ 6. 11.5] [ 5. 12.5] [ 5. 13.5] [ 5. 14.5] [ 5. 15.5] [ 5. 16.5] [ 5. 17.5] [ 5. 18.5] [ 5. 19.5] [ 5. 20.5] [ 5. 21.5] [ 5. 22.5] [ 5. 23.5] [ 5. 24.5] [ 6. 25.5] [ 5. 24.5] [ 6. 25.5] [ 7. 25.5] [ 8. 23.5] [ 9. 20.5] [ 10. 16.5] [ 9. 4.5] [ 10. 5.5] [ 10. 6.5] [ 10. 7.5] [ 10. 8.5] [ 10. 9.5] [ 10. 10.5] [ 10. 11.5] [ 10. 12.5] [ 10. 13.5] [ 10. 14.5] [ 10. 15.5] [ 10. 16.5] [ 9. 17.5] [ 9. 18.5] [ 9. 19.5] [ 9. 20.5] [ 8. 21.5] [ 8. 22.5] [ 8. 23.5] [ 7. 24.5] [ 7. 25.5] [ 5. 13.5] [ 6. 7.5] [ 7. 5.5] [ 8. 5.5] [ 9. 5.5] [ 10. 6.5] [ 7. 5.5] [ 7. 6.5] [ 6. 7.5] [ 6. 8.5] [ 6. 9.5] [ 6. 10.5] [ 6. 11.5] [ 6. 12.5] [ 5. 13.5] [ 5. 14.5] [ 5. 15.5] [ 5. 16.5] [ 5. 17.5] [ 5. 18.5] [ 5. 19.5] [ 5. 20.5] [ 5. 21.5] [ 5. 22.5] [ 5. 23.5] [ 5. 24.5] [ 5. 25.5] [ 6. 26.5] [ 5. 25.5] [ 6. 26.5] [ 7. 26.5] [ 8. 24.5] [ 9. 21.5] [ 10. 17.5] [ 9. 5.5] [ 10. 6.5] [ 10. 7.5] [ 10. 8.5] [ 10. 9.5] [ 10. 10.5] [ 10. 11.5] [ 10. 12.5] [ 10. 13.5] [ 10. 14.5] [ 10. 15.5] [ 10. 16.5] [ 10. 17.5] [ 9. 18.5] [ 9. 19.5] [ 9. 20.5] [ 9. 21.5] [ 8. 22.5] [ 8. 23.5] [ 8. 24.5] [ 7. 25.5] [ 7. 26.5] [ 4.5 13. ] [ 5.5 7. ] [ 6.5 5. ] [ 7.5 5. ] [ 8.5 5. ] [ 9.5 6. ] [ 6.5 5. ] [ 6.5 6. ] [ 5.5 7. ] [ 5.5 8. ] [ 5.5 9. ] [ 5.5 10. ] [ 5.5 11. ] [ 5.5 12. ] [ 4.5 13. ] [ 4.5 14. ] [ 4.5 15. ] [ 4.5 16. ] [ 4.5 17. ] [ 4.5 18. ] [ 4.5 19. ] [ 4.5 20. ] [ 4.5 21. ] [ 4.5 22. ] [ 4.5 23. ] [ 4.5 24. ] [ 4.5 25. ] [ 5.5 26. ] [ 4.5 25. ] [ 5.5 26. ] [ 6.5 26. ] [ 7.5 24. ] [ 8.5 21. ] [ 9.5 17. ] [ 8.5 5. ] [ 9.5 6. ] [ 9.5 7. ] [ 9.5 8. ] [ 9.5 9. ] [ 9.5 10. ] [ 9.5 11. ] [ 9.5 12. ] [ 9.5 13. ] [ 9.5 14. ] [ 9.5 15. ] [ 9.5 16. ] [ 9.5 17. ] [ 8.5 18. ] [ 8.5 19. ] [ 8.5 20. ] [ 8.5 21. ] [ 7.5 22. ] [ 7.5 23. ] [ 7.5 24. ] [ 6.5 25. ] [ 6.5 26. ] [ 5.5 13. ] [ 6.5 7. ] [ 7.5 5. ] [ 8.5 5. ] [ 9.5 5. ] [ 10.5 6. ] [ 7.5 5. ] [ 7.5 6. ] [ 6.5 7. ] [ 6.5 8. ] [ 6.5 9. ] [ 6.5 10. ] [ 6.5 11. ] [ 6.5 12. ] [ 5.5 13. ] [ 5.5 14. ] [ 5.5 15. ] [ 5.5 16. ] [ 5.5 17. ] [ 5.5 18. ] [ 5.5 19. ] [ 5.5 20. ] [ 5.5 21. ] [ 5.5 22. ] [ 5.5 23. ] [ 5.5 24. ] [ 5.5 25. ] [ 6.5 26. ] [ 5.5 25. ] [ 6.5 26. ] [ 7.5 26. ] [ 8.5 24. ] [ 9.5 21. ] [ 10.5 17. ] [ 9.5 5. ] [ 10.5 6. ] [ 10.5 7. ] [ 10.5 8. ] [ 10.5 9. ] [ 10.5 10. ] [ 10.5 11. ] [ 10.5 12. ] [ 10.5 13. ] [ 10.5 14. ] [ 10.5 15. ] [ 10.5 16. ] [ 10.5 17. ] [ 9.5 18. ] [ 9.5 19. ] [ 9.5 20. ] [ 9.5 21. ] [ 8.5 22. ] [ 8.5 23. ] [ 8.5 24. ] [ 7.5 25. ] [ 7.5 26. ]] QH6228 Qhull internal error (qh_findbestlower): all neighbors of facet 12 are flipped or upper Delaunay. Please report this error to qhull_bug@qhull.org with the input and all of the output. ERRONEOUS FACET: - f12 - flags: top simplicial upperDelaunay - normal: -0.965 -0.1942 0.1764 - offset: 6.243641 - vertices: p138(v7) p224(v3) p112(v0) - neighboring facets: f6 f28 f13

While executing: | qhull d Qz Qbb Qt Options selected for Qhull 2010.1 2010/01/14: run-id 1565793501 delaunay Qz-infinity-point Qbbound-last Qtriangulate _pre-merge _zero-centrum Pgood _max-width 22 Error-roundoff 3.7e-14 _one-merge 2.6e-13 _near-inside 1.3e-12 Visible-distance 7.4e-14 U-coplanar-distance 7.4e-14 Width-outside 1.5e-13 _wide-facet 4.4e-13 Last point added to hull was p132.

At error exit:

Delaunay triangulation by the convex hull of 225 points in 3-d:

Number of input sites and at-infinity: 11 Number of nearly incident points: 9 Number of Delaunay regions: 0

Statistics for: | qhull d Qz Qbb Qt

Number of points processed: 11 Number of hyperplanes created: 31 Number of facets in hull: 18 Number of distance tests for qhull: 1324

precision problems (corrected unless 'Q0' or an error) 8 coplanar points during partitioning