Mailman 3 March 2015 - scikit-image

Re: Hessian-Laplace blob detector.
by Vighnesh Birodkar March 9, 2015

March 9, 2015

Hey Daniil A good technical discussion is always welcome. Its never considered as spamming. Thanks Vighnesh On Monday, March 9, 2015 at 6:29:28 AM UTC+5:30, Daniil Pakhomov wrote: > > Thanks. > I've sent it. > > 2015-03-09 1:39 GMT+01:00 Josh Warner <silvertr...(a)gmail.com <javascript:> > >: > >> We'd welcome this as a PR on GitHub. That would be the ideal place for >> code review, etc. >> >> On Sunday, March 8, 2015 at 4:01:12 PM UTC-5, Daniil Pakhomov wrote: >>> >>> Really sorry for spamming you with questions. >>> No more need to answer. >>> I implemented this detector and it works as fast as your determinant of >>> Hessian approach implementation. >>> It passes all you tests and works better with coin() images (it doesn't >>> detect a false coin as determinant of Hessian does in the example). >>> >>> May I ask you to do a review of my code later? >>> >>> Thank you. >>> >>> понедельник, 2 марта 2015 г., 18:05:09 UTC+1 пользователь Daniil >>> Pakhomov написал: >>>> >>>> Hello, >>>> >>>> I want to try to implement Hessian-Laplace blob detector (as mentioned >>>> in requested features on github page). >>>> >>>> Can someone give me the list of corresponding papers, using which I can >>>> implement it. >>>> >>>> Thank you. >>>> >>> -- >> You received this message because you are subscribed to a topic in the >> Google Groups "scikit-image" group. >> To unsubscribe from this topic, visit >> https://groups.google.com/d/topic/scikit-image/ghIYwQFubEU/unsubscribe. >> To unsubscribe from this group and all its topics, send an email to >> scikit-image...(a)googlegroups.com <javascript:>. >> For more options, visit https://groups.google.com/d/optout. >> > >

1 0

Hessian-Laplace blob detector.
by Daniil Pakhomov March 9, 2015

March 9, 2015

Hello, I want to try to implement Hessian-Laplace blob detector (as mentioned in requested features on github page). Can someone give me the list of corresponding papers, using which I can implement it. Thank you.

4 6

ANN: scikit-image 0.11
by Stefan van der Walt March 5, 2015

March 5, 2015

Announcement: scikit-image 0.11.0 ================================= We're happy to announce the release of scikit-image v0.11.0! scikit-image is an image processing toolbox for SciPy that includes algorithms for segmentation, geometric transformations, color space manipulation, analysis, filtering, morphology, feature detection, and more. For more information, examples, and documentation, please visit our website: http://scikit-image.org Highlights ---------- For this release, we merged over 200 pull requests with bug fixes, cleanups, improved documentation and new features. Highlights include: - Region Adjacency Graphs - Color distance RAGs (#1031) - Threshold Cut on RAGs (#1031) - Similarity RAGs (#1080) - Normalized Cut on RAGs (#1080) - RAG drawing (#1087) - Hierarchical merging (#1100) - Sub-pixel shift registration (#1066) - Non-local means denoising (#874) - Sliding window histogram (#1127) - More illuminants in color conversion (#1130) - Handling of CMYK images (#1360) - `stop_probability` for RANSAC (#1176) - Li thresholding (#1376) - Signed edge operators (#1240) - Full ndarray support for `peak_local_max` (#1355) - Improve conditioning of geometric transformations (#1319) - Standardize handling of multi-image files (#1200) - Ellipse structuring element (#1298) - Multi-line drawing tool (#1065), line handle style (#1179) - Point in polygon testing (#1123) - Rotation around a specified center (#1168) - Add `shape` option to drawing functions (#1222) - Faster regionprops (#1351) - `skimage.future` package (#1365) - More robust I/O module (#1189) API Changes ----------- - The ``skimage.filter`` subpackage has been renamed to ``skimage.filters``. - Some edge detectors returned values greater than 1--their results are now appropriately scaled with a factor of ``sqrt(2)``. Contributors to this release ---------------------------- (Listed alphabetically by last name) - Fedor Baart - Vighnesh Birodkar - François Boulogne - Nelson Brown - Alexey Buzmakov - Julien Coste - Phil Elson - Adam Feuer - Jim Fienup - Geoffrey French - Emmanuelle Gouillart - Charles Harris - Jonathan Helmus - Alexander Iacchetta - Ivana Kajić - Kevin Keraudren - Almar Klein - Gregory R. Lee - Jeremy Metz - Stuart Mumford - Damian Nadales - Pablo Márquez Neila - Juan Nunez-Iglesias - Rebecca Roisin - Jasper St. Pierre - Jacopo Sabbatini - Michael Sarahan - Salvatore Scaramuzzino - Phil Schaf - Johannes Schönberger - Tim Seifert - Arve Seljebu - Steven Silvester - Julian Taylor - Matěj Týč - Alexey Umnov - Pratap Vardhan - Stefan van der Walt - Joshua Warner - Tony S Yu

2 1

GSoC - Dynamic Time Warping
by Mudit Jain March 5, 2015

March 5, 2015

Hello everyone, I am Mudit, a third year undergraduate student from Birla Institute of Technology and Sciences, India. I had come across the idea of implementing a Dynamic Time Warping Library on the Ideas Page and would like to pursue the project for GSOC-2015. I have read the paper regarding Adaptive Feature Based Dynamic Time Warping Library, that was attached in the Ideas page and have a thorough understanding of the system to be implemented. I have attached a summary regarding the same. I have also given a procedure to implement the methodology stated in the paper. I understand that the implementation of dynamic programming is similar to that of the edit distance problem.The summary also has a description of the workflow for the algorithm used. I am presently going through the DTW package is R Some links that I am following are: http://www.jstatsoft.org/v31/i07/paper It would be highly appreciated if more insight regarding the project and what would be the role of a person if he/she is selected for this project. I would also appreciate links that can better my understanding regarding various topics. Cheers Mudit

1 0

Re: hough ellipse fit inaccurate?
by Kevin Keraudren March 5, 2015

March 5, 2015

By the way, your ground truth is the values you used when drawing the ellipse, not the values you detect with a different detection method. On Thu, Mar 5, 2015 at 10:07 AM, Kevin Keraudren < kevin.keraudren(a)googlemail.com> wrote: > Hi Arno, > In order to stay on the safe side, why don't you post your actual code, > with a test case highlighting the error you measure between the true centre > of the ellipse and the detected one? > Kind regards, > Kevin > > On Thu, Mar 5, 2015 at 9:53 AM, Arno Dietz <arnodietz86(a)googlemail.com> > wrote: > >> Thank you very much Kevin and Johannes. >> >> I see the rounding Problem, but it is just for the ellipse drawing. In my >> actual code I just use the Ellipse center like best [1] and best[2] without >> rounding. This still produces much more inaccurate ellipse center results >> than other methods like center of mass for example, althoug >> <http://www.dict.cc/englisch-deutsch/although.html>h I also use >> the anti-aliased input image. So is there any possibility to get more >> accurate results from the hough ellipse fit approach? If not, this is also >> ok, I just want to be on the safe side that it's not my fault. In that >> case I will have a look at the suggested approach from Johannes. >> >> Am Donnerstag, 5. März 2015 01:21:48 UTC+1 schrieb Kevin Keraudren: >>> >>> Hi Arno, >>> >>> The first source of inaccuracy comes from your code, you need to round >>> the values instead of truncating them: >>> >>> #yc = int(best[1]) >>> >>> >>> #xc = int(best[2]) >>> >>> >>> #a = int(best[3]) >>> >>> >>> #b = int(best[4]) >>> >>> >>> >>> yc = int(round(best[1])) >>> >>> xc = int(round(best[2])) >>> >>> a = int(round(best[3])) >>> >>> b = int(round(best[4])) >>> >>> See resulting image attached. >>> >>> Kind regards, >>> >>> Kevin >>> >>> >>> A second source of inaccuracy comes from your input ellipse: it is not a >>> perfect ellipse because you drew it using anti-aliasing. >> >> >> >> Third, you could fit an ellipse using RANSAC. How does this approach work >>> for you: http://stackoverflow.com/questions/28281742/fitting-a- >>> circle-to-a-binary-image/28289147#28289147 >> >> -- >> You received this message because you are subscribed to the Google Groups >> "scikit-image" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to scikit-image+unsubscribe(a)googlegroups.com. >> For more options, visit https://groups.google.com/d/optout. >> > >

1 0

Re: hough ellipse fit inaccurate?
by Arno Dietz March 5, 2015

March 5, 2015

Thank you very much Kevin and Johannes. I see the rounding Problem, but it is just for the ellipse drawing. In my actual code I just use the Ellipse center like best [1] and best[2] without rounding. This still produces much more inaccurate ellipse center results than other methods like center of mass for example, althoug <http://www.dict.cc/englisch-deutsch/although.html>h I also use the anti-aliased input image. So is there any possibility to get more accurate results from the hough ellipse fit approach? If not, this is also ok, I just want to be on the safe side that it's not my fault. In that case I will have a look at the suggested approach from Johannes. Am Donnerstag, 5. März 2015 01:21:48 UTC+1 schrieb Kevin Keraudren: > > Hi Arno, > > The first source of inaccuracy comes from your code, you need to round the > values instead of truncating them: > > #yc = int(best[1]) > > > #xc = int(best[2]) > > > #a = int(best[3]) > > > #b = int(best[4]) > > > > yc = int(round(best[1])) > > xc = int(round(best[2])) > > a = int(round(best[3])) > > b = int(round(best[4])) > > See resulting image attached. > > Kind regards, > > Kevin > > > A second source of inaccuracy comes from your input ellipse: it is not a > perfect ellipse because you drew it using anti-aliasing. Third, you could fit an ellipse using RANSAC. How does this approach work > for you: > http://stackoverflow.com/questions/28281742/fitting-a-circle-to-a-binary-im… >

2 1

Re: hough ellipse fit inaccurate?
by Johannes Schoenberger March 5, 2015

March 5, 2015

Third, you could fit an ellipse using RANSAC. How does this approach work for you: http://stackoverflow.com/questions/28281742/fitting-a-circle-to-a-binary-im… > On Mar 4, 2015, at 7:24 PM, Kevin Keraudren <kevin.keraudren(a)googlemail.com> wrote: > > A second source of inaccuracy comes from your input ellipse: it is not a perfect ellipse because you drew it using anti-aliasing. > > On Thu, Mar 5, 2015 at 12:21 AM, Kevin Keraudren <kevin.keraudren(a)googlemail.com> wrote: > Hi Arno, > > The first source of inaccuracy comes from your code, you need to round the values instead of truncating them: > > #yc = int(best[1]) > #xc = int(best[2]) > #a = int(best[3]) > #b = int(best[4]) > > yc = int(round(best[1])) > xc = int(round(best[2])) > a = int(round(best[3])) > b = int(round(best[4])) > > See resulting image attached. > > Kind regards, > > Kevin > > > > On Wed, Mar 4, 2015 at 11:49 PM, Arno Dietz <arnodietz86(a)googlemail.com> wrote: > > Ok sorry. Here is my code: > > from skimage import color > from skimage.filter import canny > from skimage.transform import hough_ellipse > from skimage.draw import ellipse_perimeter > from skimage import io > from skimage.viewer import ImageViewer > # load image > img = io.imread('ellipse.png') > cimg = color.gray2rgb(img) > # edges and ellipse fit > edges = canny(img, sigma=0.1, low_threshold=0.55, high_threshold=0.8) > result = hough_ellipse(edges, accuracy=4, threshold=25, min_size=47, max_size=60) > result.sort(order='accumulator') > # Estimated parameters for the ellipse > best = result[-1] > yc = int(best[1]) > xc = int(best[2]) > a = int(best[3]) > b = int(best[4]) > orientation = best[5] > # Draw the ellipse on the original image > cy, cx = ellipse_perimeter(yc, xc, a, b, orientation) > cimg[cy, cx] = (0, 0, 255) > # Draw the edge (white) and the resulting ellipse (red) > edges = color.gray2rgb(edges) > edges[cy, cx] = (250, 0, 0) > viewer = ImageViewer(edges) > viewer.show() > > I noticed, that the ellipse center is detected only in half pixel accuracy. Maybe this is the Problem? Is there a possibility to get the ellipse center with sub-pixel accuracy? > > regards Arno > > -- > You received this message because you are subscribed to the Google Groups "scikit-image" group. > To unsubscribe from this group and stop receiving emails from it, send an email to scikit-image+unsubscribe(a)googlegroups.com. > For more options, visit https://groups.google.com/d/optout. > > > > -- > You received this message because you are subscribed to the Google Groups "scikit-image" group. > To unsubscribe from this group and stop receiving emails from it, send an email to scikit-image+unsubscribe(a)googlegroups.com. > For more options, visit https://groups.google.com/d/optout.

1 0

Re: hough ellipse fit inaccurate?
by Kevin Keraudren March 5, 2015

March 5, 2015

A second source of inaccuracy comes from your input ellipse: it is not a perfect ellipse because you drew it using anti-aliasing. On Thu, Mar 5, 2015 at 12:21 AM, Kevin Keraudren < kevin.keraudren(a)googlemail.com> wrote: > Hi Arno, > > The first source of inaccuracy comes from your code, you need to round the > values instead of truncating them: > > #yc = int(best[1]) > > > #xc = int(best[2]) > > > #a = int(best[3]) > > > #b = int(best[4]) > > > > yc = int(round(best[1])) > > xc = int(round(best[2])) > > a = int(round(best[3])) > > b = int(round(best[4])) > > See resulting image attached. > > Kind regards, > > Kevin > > > > On Wed, Mar 4, 2015 at 11:49 PM, Arno Dietz <arnodietz86(a)googlemail.com> > wrote: > >> >> Ok sorry. Here is my code: >> >> from skimage import color >>> from skimage.filter import canny >>> from skimage.transform import hough_ellipse >>> from skimage.draw import ellipse_perimeter >>> from skimage import io >>> from skimage.viewer import ImageViewer >>> # load image >>> img = io.imread('ellipse.png') >>> cimg = color.gray2rgb(img) >>> # edges and ellipse fit >>> edges = canny(img, sigma=0.1, low_threshold=0.55, high_threshold=0.8) >>> result = hough_ellipse(edges, accuracy=4, threshold=25, min_size=47, >>> max_size=60) >>> result.sort(order='accumulator') >>> # Estimated parameters for the ellipse >>> best = result[-1] >>> yc = int(best[1]) >>> xc = int(best[2]) >>> a = int(best[3]) >>> b = int(best[4]) >>> orientation = best[5] >>> # Draw the ellipse on the original image >>> cy, cx = ellipse_perimeter(yc, xc, a, b, orientation) >>> cimg[cy, cx] = (0, 0, 255) >>> # Draw the edge (white) and the resulting ellipse (red) >>> edges = color.gray2rgb(edges) >>> edges[cy, cx] = (250, 0, 0) >>> viewer = ImageViewer(edges) >>> viewer.show() >> >> >> I noticed, that the ellipse center is detected only in half pixel >> accuracy. Maybe this is the Problem? Is there a possibility to get the >> ellipse center with sub-pixel accuracy? >> >> regards Arno >> >>> -- >> You received this message because you are subscribed to the Google Groups >> "scikit-image" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to scikit-image+unsubscribe(a)googlegroups.com. >> For more options, visit https://groups.google.com/d/optout. >> > >

1 0

hough ellipse fit inaccurate?
by Arno Dietz March 5, 2015

March 5, 2015

3 3

Re: GSoC 2015 mentors
by Johannes Schoenberger March 4, 2015

March 4, 2015

Done. > On Mar 4, 2015, at 5:56 PM, Stéfan van der Walt <stefanv(a)berkeley.edu> wrote: > > On Wed, Mar 4, 2015 at 2:21 PM, Johannes Schoenberger <jsch(a)demuc.de> wrote: > Sorry, I’ll be too busy over the summer myself, but I am happy to function as a backup mentor, if occasional reviews of PRs and sporadic participation in discussions are okay with you. > > Good enough! Please fill out the mentor signup form here: > > http://goo.gl/forms/PMXVM1CUAS > > Stéfan > > -- > You received this message because you are subscribed to the Google Groups "scikit-image" group. > To unsubscribe from this group and stop receiving emails from it, send an email to scikit-image+unsubscribe(a)googlegroups.com. > For more options, visit https://groups.google.com/d/optout.

1 0