Mailman 3 June 2016 - scikit-image

Re: FW: Python version of your subpixel registration code
by Josh Warner 18 Jun '16

18 Jun '16

Tangentially related: I'd very much like to see a highly accurate rotational/angular registration algorithm added to scikit-image to complement this one. From my own experience, when combining a bracketed set into HDR images the alignment must be near-perfect. This does a great job for translation, but often there's a rotational component also. I've got another great advanced tutorial example brewing, but need highly accurate rotational registration to make it really properly robust... Josh On Friday, June 17, 2016 at 2:47:33 AM UTC-5, stefanv wrote: > > Hi Michael > > Thanks for letting us know--as you correctly point out, our code was > vetted & approved by the original authors, so we are not affected. > > Best regards > Stéfan > > On 16 June 2016 at 05:44, Michael Sarahan <msarahan(a)gmail.com> wrote: > >> Manuel, >> >> Many thanks for the good news! I have copied the scikit image mailing >> list here. Stefan was able to get permission from Prof. Fienup so that we >> could include it, but this news is still excellent. Scikit image team, >> please ping me on github if this news raises any issues. >> >> Best regards, >> Michael >> >> On Thu, Jun 16, 2016, 04:21 Manuel Guizar <mguizar(a)gmail.com> wrote: >> >>> Hi Michael, >>> >>> I am not sure if you are still interested or if this has any impact on >>> your scikit contribution. Took some time but now the code is under a BSD >>> license. I had to rewrite parts of it in the end. >>> >>> >>> http://ch.mathworks.com/matlabcentral/fileexchange/18401-efficient-subpixel… >>> >>> Cheers, >>> >>> >>> -- >>> Manuel Guizar-Sicairos, PhD >>> >> -- >> 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. >> To post to this group, send email to scikit-image(a)googlegroups.com. >> To view this discussion on the web, visit >> https://groups.google.com/d/msgid/scikit-image/CAB9hrOqzb0_Uo5xTvWSnaxoXRQB… >> <https://groups.google.com/d/msgid/scikit-image/CAB9hrOqzb0_Uo5xTvWSnaxoXRQB…> >> . >> For more options, visit https://groups.google.com/d/optout. >> > >

1 0

Re: FW: Python version of your subpixel registration code
by Stéfan van der Walt 17 Jun '16

17 Jun '16

Hi Michael Thanks for letting us know--as you correctly point out, our code was vetted & approved by the original authors, so we are not affected. Best regards Stéfan On 16 June 2016 at 05:44, Michael Sarahan <msarahan(a)gmail.com> wrote: > Manuel, > > Many thanks for the good news! I have copied the scikit image mailing list > here. Stefan was able to get permission from Prof. Fienup so that we > could include it, but this news is still excellent. Scikit image team, > please ping me on github if this news raises any issues. > > Best regards, > Michael > > On Thu, Jun 16, 2016, 04:21 Manuel Guizar <mguizar(a)gmail.com> wrote: > >> Hi Michael, >> >> I am not sure if you are still interested or if this has any impact on >> your scikit contribution. Took some time but now the code is under a BSD >> license. I had to rewrite parts of it in the end. >> >> >> http://ch.mathworks.com/matlabcentral/fileexchange/18401-efficient-subpixel… >> >> Cheers, >> >> >> -- >> Manuel Guizar-Sicairos, PhD >> > -- > 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. > To post to this group, send email to scikit-image(a)googlegroups.com. > To view this discussion on the web, visit > https://groups.google.com/d/msgid/scikit-image/CAB9hrOqzb0_Uo5xTvWSnaxoXRQB… > <https://groups.google.com/d/msgid/scikit-image/CAB9hrOqzb0_Uo5xTvWSnaxoXRQB…> > . > For more options, visit https://groups.google.com/d/optout. >

1 0

Re: Thresholding a dark region in an image
by François Boulogne 17 Jun '16

17 Jun '16

Le 16/06/2016 à 12:45, Nabil Freij a écrit : > Hi Juan and François, > > Thanks for those suggestions. So from my > understanding, threshold_isodata, threshold_li, threshold_otsu and > threshold_yen return values that allow me to threshold an image. Using > these on the whole image returns a threshold value which return a > value which does a good job of getting the entire sunspot. Which can > be seen in Figure 1. Varying the bin size, where possible, changes > very little. So I decided to use them on a smaller image that shows > just the sunspot and the results are very good as you can see in > Figure 2. I will carry on trying the various routines to see what the > results are. > > I used the minimum algorithm when it was merged into scikit master and > I have found it to be the most useful in my cases so far. But my > current results are very rough. > You can give a try to multiotsu. https://github.com/scikit-image/scikit-image/pull/2076 It's not merged yet, there is a missing import (from skimage.util import img_as_float). For the rest, it should be good enough for a quick test. :) Feel free to comment on this one. -- François Boulogne. http://www.sciunto.org GPG: 32D5F22F

1 0

Re: FW: Python version of your subpixel registration code
by Michael Sarahan 16 Jun '16

16 Jun '16

Manuel, Many thanks for the good news! I have copied the scikit image mailing list here. Stefan was able to get permission from Prof. Fienup so that we could include it, but this news is still excellent. Scikit image team, please ping me on github if this news raises any issues. Best regards, Michael On Thu, Jun 16, 2016, 04:21 Manuel Guizar <mguizar(a)gmail.com> wrote: > Hi Michael, > > I am not sure if you are still interested or if this has any impact on > your scikit contribution. Took some time but now the code is under a BSD > license. I had to rewrite parts of it in the end. > > > http://ch.mathworks.com/matlabcentral/fileexchange/18401-efficient-subpixel… > > Cheers, > > > -- > Manuel Guizar-Sicairos, PhD >

1 0

Re: Thresholding a dark region in an image
by Juan Nunez-Iglesias 16 Jun '16

16 Jun '16

Hi Nabil, Have you looked at the skimage.filters.threshold_* functions? I would start by looking at the results from each of those on a few example sunspots to see whether one reliably gives you a reasonable result. Juan. On 20 May 2016 at 12:26:37 AM, Nabil Freij (nabil.freij(a)gmail.com) wrote: Hello, I have been investigating how to threshold a structure in my images. More specifically, a sunspot, that can be seen in Figure 1. The aim has been to isolate and measure the number of pixels that are part of the central part of a sunspot, which is the darkest region of the structure. The method I have been using is to define a region which contains no features that are similar to a sunspot. The size of this region is not fixed but I generally make it as large as the data allows. This region is at the bottom black box in Figure 1 (ignore the slider). The mean and standard deviation of this region is calculated, which allows me to define a threshold limit, which is subtracting the standard deviation multiplied by a user defined constant from the mean. The constant is chosen so that overall, the threshold is selecting the pixels we know that are part of the central region. But this value varies depending the structure and the data source. Another method I have discussed or looked into are related to Figure 2. On the left is a cropped field of view of the sunspot and on the right is a histogram of this image in red and in yellow is the histogram of the background box from Figure 1. By using the histogram, I can work out the numerical gradient and pick the points with the largest shift. I also tried this method on the original image by taking slices along the sunspot, however, due to the non-uniform nature of a sunspot the results were not very good. I was wondering if there were any suggestions to threshold this region that is not as ad hoc as my method? Thanks, Nabil -- 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. To post to this group, send email to scikit-image(a)googlegroups.com. To view this discussion on the web, visit https://groups.google.com/d/msgid/scikit-image/CADwJ0z1Sthz7OW8uu2G2i%2Bm9V…. For more options, visit https://groups.google.com/d/optout.

2 1

Testing of object detection module.
by Daniil Pakhomov 16 Jun '16

16 Jun '16

Hello everyone. I am testing an object detection module in scikit-image. Currently I have added a face detection file. If someone is interested and have time to test it on their PC, it would be great. Just clone this branch: https://github.com/warmspringwinds/scikit-image/tree/object_detect_module and run cython compilation in the root of the repository: python setup.py build_ext --inplace After that run the object_detection.ipynb <https://github.com/warmspringwinds/scikit-image/blob/object_detect_module/o…> ipython notebook file and see if it works for you. Please, report any problems that you 'face' :P I will appreciate any help :)

1 0

Re: New core team member
by Egor Panfilov 15 Jun '16

15 Jun '16

A warm welcome to you, Francois, from Eastern Europe! Regards, Egor Panfilov вторник, 14 июня 2016 г., 23:52:20 UTC+3 пользователь stefanv написал: > > Hi everyone > > I am happy to announce that François Boulogne is joining the scikit-image > core team. Welcome, François! We look forward to working with and > learning from you. > > Best regards > Stéfan > >

2 1

Interesting to include a the Scattering transform in sci-kit image?
by Sira Ferradans 15 Jun '16

15 Jun '16

Dear Sci-kit image developers, at the ENS (Paris) we are planning on implementing a Python version of the Scattering transform 2D and 1D. The scattering transform has proven to be very powerful as a descriptor for image classification and signal analysis. We thought that it may be useful to integrate the 2D version in the sci-kit image package, since it aligns well with the software package and the community. The idea would be to implement the functionalities following as much as possible the APIs you already have for similar functions. More specifically, we will need the following (approximately): -Morlet wavelet: Closely related to the Gabor wavelet, so it should take into account its API -Scattering transform: output the scattering transform coefficients, either for display or (as a vector) for learning purposes. We are attaching a small tutorial (better visualized if you download it) that compares the performance of the first order scattering coefficients computed with Gabor filters, versus the coefficients you extract in the *'Gabor filter banks for texture classification' <http://scikit-image.org/docs/dev/auto_examples/features_detection/plot_gabo…>* tutorial. The goal of this ipython notebook is to show that the implementation can be easily integrated in you library while providing a powerful tool for image analysis. If you think this is a good idea, please let us know. Moreover, it would be great if you could give us some guidelines that you think would make the process easier. We will be adhering to the instructions given on the contributions page, but please don't hesitate to give feedback on our PR! Best Regards, Michael Eickenberg and Sira Ferradans.

2 1

New core team member
by Stéfan van der Walt 15 Jun '16

15 Jun '16

Hi everyone I am happy to announce that François Boulogne is joining the scikit-image core team. Welcome, François! We look forward to working with and learning from you. Best regards Stéfan

1 0

Re: Haarcascades implementation
by Daniil Pakhomov 14 Jun '16

14 Jun '16

You are welcome. In the original paper, the authors got better results using MB-LBP than using Haar features: http://www.cbsr.ia.ac.cn/users/scliao/papers/Zhang-ICB07-MBLBP.pdf The scikit-learn wasn't used because: 1) They don't have gentle adaboost. 2) Their implementation is based on pure Python, which is slow. Overall, the implementation aims to provide very fast detection with good precision. The implementation also enables one to use OpenMP, which makes it almost real-time for some cases. воскресенье, 12 июня 2016 г., 17:54:58 UTC-7 пользователь Abdeali Kothari написал: > > Hi Daniil, > > Thank you very much for that mail > > The LBP features do look interesting, and I had no idea OpenCV had that ! > It seems LBP is considerably faster than HAAR, but not as accurate[1]. > Which makes me wonder whether that is the right choice. Also, there are > more pre-trained haarcascades than lbpcascades which makes me favor > haarcascade more. > > But this implementation is something that I can read and understand how to > do it myself better ! Thanks :) > > I'm wondering as to why sklearn was not used in pull/1570 [2] ? It seems > to me like a lot of the code in skimage/future/detect/cascade.pyx [3] would > already be available there (Probably in sklearn.ensemble.AdaBoostClassifier > [4])? It could be made an "optional" feature which only works if sklearn is > installed ? > > Regards, > Abdeali JK > > [1] - > http://stackoverflow.com/questions/8791178/haar-cascades-vs-lbp-cascades-in… > [2] - https://github.com/scikit-image/scikit-image/pull/1570 > [3] - > https://github.com/warmspringwinds/scikit-image/blob/7face0ebbfadab299b0e2a… > [4] - > http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.AdaBoostC… > > On Mon, Jun 13, 2016 at 4:53 AM, Daniil Pakhomov <daniil.j...(a)gmail.com > <javascript:>> wrote: > >> Also, this might be useful: >> http://warmspringwinds.github.io/ >> >> воскресенье, 12 июня 2016 г., 10:37:18 UTC-4 пользователь Abdeali Kothari >> написал: >>> >>> Hi, I recently switched from opencv to skimage as I found it much more >>> pythonic and easier to install. >>> >>> There is one feature which I seem to be missing - The Haarcascades and >>> CascadeClassifier[1]. Is there plan to add this in skimage ? I'm quite new >>> to haarcascades, but my understanding is that first there are HAAR features >>> that are found from the image and then these features are classified using >>> multiple weak classifiers (using adaboost) to "detect" an object. >>> >>> It seems to me that to get this functionality I would need skimage (to >>> create the haar features) and sklearn (for the adaboost setup). I found >>> issues/1431[2] which seems to talk about creating HAAR features. >>> >>> Am I on the right track ? Is there already some example code available >>> which uses skimage and sklearn using one of the haarcascadde xml files >>> provided by opencv ? >>> >>> Regards, >>> Abdeali JK >>> >>> >>> [1] - >>> http://docs.opencv.org/master/d7/d8b/tutorial_py_face_detection.html#gsc.ta… >>> [2] - https://github.com/scikit-image/scikit-image/issues/1431 >>> >> -- >> 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/PIpELINpmek/unsubscribe. >> To unsubscribe from this group and all its topics, send an email to >> scikit-image...(a)googlegroups.com <javascript:>. >> To post to this group, send email to scikit...(a)googlegroups.com >> <javascript:>. >> To view this discussion on the web, visit >> https://groups.google.com/d/msgid/scikit-image/bb52ccad-e9fe-4a61-b980-2677… >> <https://groups.google.com/d/msgid/scikit-image/bb52ccad-e9fe-4a61-b980-2677…> >> . >> >> For more options, visit https://groups.google.com/d/optout. >> > >

1 0