Mailman 3 January 2014 - scikit-image

Re: skimage and opencv
by Emmanuelle Gouillart 22 Sep '16

22 Sep '16

Hi FranÃ¯Â¿Â½ois, that's an excellent question, and not a troll :-). Opencv is a very powerful library, but it focuses primarily on computer vision (feature detection and extraction, classification, ...), as opposed to image processing in general (with other tasks such as denoising, segmentation, ...). The other big difference is that skimage builds on numpy ndarrays, and uses the full power of the numpy API (including of course the basic facilities for processing arrays as images that come with numpy), as well as some of scipy functions (you could have added scipy.ndimage to your list -- a few functions in skimage are wrappers around scipy.ndimage, that exist for the sake of completeneness). One important consequence is that algorithms working for 3-d or even n-d images can be easily implemented in 3-d/n-d in skimage, whereas opencv is restricted to 2-D images (as far as I know). Thanks to the use of numpy arrays, the API of skimage is also quite pleasant for a numpy user, more than the API of opencv. A related difference is that skimage is written in python and cython, whereas opencv is a C++ library. The two libraries attract a different crowd of developers, and a Python/Cython toolkit based on numpy arrays is easier to develop and maintain inside the Scientific Python ecosystem. I'm sure that other devs/users will have things to add to this discussion! Cheers, Emmanuelle On Thu, Dec 27, 2012 at 02:06:08PM -0800, FranÃ¯Â¿Â½ois wrote: > Hi users and devs, > It came to my knowledge that another python library (based on C++ and C > codes) for image processing exists too : opencv > I understand that numpy intregrates some basic features and we need some > advanced features but I have the feeling that skimages is redoundant with > opencv in some ways. > What's the position of skimage about that? (Don't read this question as a > troll but like a real question). > I mean that similar features exist in both. Would not be possible to > reuse/integrate opencv or merge? what's the reason for keeping them apart? > My observation is there is 4 libraries to manipulate images: > * PIL > * numpy > * skimages > * opencv > That's a lot. > Cheers,

3 2

Re: Contribution: circles and hough transform
by François 18 Jun '14

18 Jun '14

Hi, FYI, I plan to work on the detection of ellipses quite soon. Stefan indicated to me a reference which seems fairly simple to implement. Cheers, François.

3 3

Re: incoherent face orientation in marching cubes
by Josh Warner 09 Feb '14

09 Feb '14

Calculate the gradient for each face, normalize it, and orient the face counterclockwise around the descent of said gradient vector. It would require some pretty heavy modification, but might be worth it. On the other hand, you could potentially hack in a fix (depends on how the STL package works) much easier by making the output double sided. A double-sided mesh has each face twice, covering both orientations This way you'd just append `faces` to itself, reverse indexing to flip the orientation of each face, e.g. ``` python import numpy as np from skimage.measure import marching_cubes verts, faces = marching_cubes(vol, value) faces = np.concatenate((faces, faces[:, -1], axis=0) # Remainder of STL export code for faces here ``` Which would be *far, far* simpler and less computationally intense than dealing with the gradients... Would you mind trying this quick fix, Janto? On Wednesday, January 8, 2014 6:35:28 PM UTC-6, Stefan van der Walt wrote: > > Josh, you said you had some ideas on how to add this--could you expand? >

3 4

Re: Non-local means denoising
by Emmanuelle Gouillart 02 Feb '14

02 Feb '14

Hi Juan, yes, it's been a long time :-) (I took a new job in September and it's been pretty hectic since then!). My implementation is indeed 3-D. I'll start working on a PR this week-end then! For 3-D images the code is pretty slow due to the large size of the data, but I also have a pycuda + cuda implementation written by my PhD student William. You'll tell me if you're interested! Cheers, Emmanuelle On Fri, Jan 24, 2014 at 11:46:19AM +1100, Juan Nunez-Iglesias wrote: > Haha, hi Emmanuelle, long time no see! ;) I think your interjection in this > thread is rather priceless. =D Is your implementation 3D? Either way, if you > submit a WIP PR, I might be tempted to use some cycles to help out. =) > On Fri, Jan 24, 2014 at 3:26 AM, Emmanuelle Gouillart < > emmanuelle.gouillart(a)nsup.org> wrote: > Hi, > it happens that I have an implementation of NL-means (implemented in C + > cython). I have to clean it a little and then I can send it to the list. > It would actually be interesting to convert it to pure Cython for further > maintanability. > Cheers, > Emmanuelle > On Thu, Jan 23, 2014 at 03:48:41PM +0000, Kevin Keraudren wrote: > > I just listed these implementations as I recently searched for a C++ > > implementation. > > Unfortunately, I cannot volunteer to work on this and I do not know > > anyone who already worked on this topic... > > Kind regards, > > Kevin > > Le 23/01/2014 11:55, StÃ©fan van der Walt a Ã©crit : > > >Hi Kevin > > >On Thu, 23 Jan 2014 09:47:44 +0000, Kevin Keraudren wrote: > > >>Here are some implementations of Non-Local Means Denoising you might > > >>be interested in for reference/comparison: > > >Thanks for these informative links. Â Are you interested in working on > nlmeans, > > >or do you know of someone who would be able to help us review the code > once it > > >is written? > > >Thanks > > >StÃ©fan

2 2

Help with using filter.rank.median
by Aman Thakral 01 Feb '14

01 Feb '14

Hi all, I'm currently trying to use a median filter on my data, an am getting an error that I can't seem to debug. I'm taking the difference of two images (both numpy.float64), scaling them between -1 and 1, and trying to apply a median filter. My image satisfies the conditions for being between -1 and 1 when I do the following test: print diff.min()/dmax < -1.0 -> returns False (LHS evals to -1.0) print diff.max()/dmax > 1.0 -> returns False (LHS evals to 0.663919388648) The statement causing the error: filtered = fltr.median(diff/dmax, selem) where diff is my difference, and dmax is my scaling adjustment, selem is a disk of radius 2 This gives me the following traceback. Traceback (most recent call last): File "/home/clustering.py", line 138, in image_ratio filtered = fltr.median(diff/dmax/2.+0.5, selem, mask=m.astype(np.bool)) File "/usr/local/lib/python2.7/dist-packages/skimage/filter/rank/generic.py", line 369, in median out=out, mask=mask, shift_x=shift_x, shift_y=shift_y) File "/usr/local/lib/python2.7/dist-packages/skimage/filter/rank/generic.py", line 71, in _apply out_dtype) File "/usr/local/lib/python2.7/dist-packages/skimage/filter/rank/generic.py", line 34, in _handle_input image = img_as_ubyte(image) File "/usr/local/lib/python2.7/dist-packages/skimage/util/dtype.py", line 358, in img_as_ubyte return convert(image, np.uint8, force_copy) File "/usr/local/lib/python2.7/dist-packages/skimage/util/dtype.py", line 191, in convert raise ValueError("Images of type float must be between -1 and 1.") ValueError: Images of type float must be between -1 and 1. Has anyone else experienced this problem before? I tried searching in the groups, but couldn't find anything relevant. Any help would be greatly appreciated! Thanks, Aman

2 1

Re: Announcing an skimage-inspired project
by Stéfan van der Walt 30 Jan '14

30 Jan '14

Hi Adam On Thu, 30 Jan 2014 03:10:30 -0500, Adam Hughes wrote: > Well, without further ado, here is the project, *pyparty. *Thanks! > > https://github.com/hugadams/pyparty This looks great! I can see you've put a lot of effort into it, especially those notebooks. The only recommendation I can make is that you make your first example one where you attack a real world particle problem. Keep up the good work! Cheers StÃ©fan

2 1

Announcing an skimage-inspired project
by Adam Hughes 30 Jan '14

30 Jan '14

Hey everyone, I have been working on a small project over the last month based on an idea posted to the list last month: https://groups.google.com/forum/#!topic/scikit-image/6LgfQLO4Jcc In my research, we do a lot of processing of electron microscope images of nanoparticles. While image processing is very simple in scikit-image (and not-so-simple but still manageable when we were writing ImageJ Python spaghetti code way back when), the post-processing always turned ad-hoc once we started to quantify, manage and manipulate the segmented particles. Simple masking worked to an extent, but got hard to juggle for all of the various things we needed to do. When I set out to tackle the "pattening/drawing" issue linked above, it occurred to me that a fairly pythonic API for "managing particles" could be adapted to particles returned by skimage.draw(), as well as skimage.morphology.labels()/descriptors(). In other words, "particles" from labeled images as well as "drawn particles" could be abstracted into a single API. This was, in a sense, the key insight to motivate the pursuit, and I think it paid off. The project is only possible because skimage.draw/labels() are so intelligently built. Except for maybe the gridding capabilities, all of the core structures are built primarily to streamline common processing operations and are designed to work for scikit-image, not in opposition to it. If any of the descriptions of notebook tutorials suggest otherwise, please let me know so that I may reword them. Well, without further ado, here is the project, *pyparty. *Thanks! https://github.com/hugadams/pyparty https://pypi.python.org/pypi

1 0

Re: scikit-image paper
by Juan Nunez-Iglesias 28 Jan '14

28 Jan '14

On Tue, Jan 28, 2014 at 11:48 PM, Juan Nunez-Iglesias <jni.soma(a)gmail.com>wrote: > Also, they apparently accept latex submissions, even if they don't provide > a template. > Aaaaand, on cue: http://blog.peerj.com/post/74820148558/peerj-partners-with-writelatex

1 0

Re: Examples relicensing proposal
by James Turner 28 Jan '14

28 Jan '14

I suppose you could do like Google etc. and provide a means for people to opt out if they don't like their contributions being distributed under CC? I'm not sure how legally sound it is in theory but I'd think it ought to pass any test of reasonableness and it shifts the inertia to where it belongs :-). Cheers, James. On 28/01/14 13:54, StÃ©fan van der Walt wrote: > On Tue, Jan 28, 2014 at 1:42 PM, Juan Nunez-Iglesias <jni.soma(a)gmail.com> wrote: >> Having read CC0, I'd be ok with the examples being under that license. I'll >> let you decide whether you prefer the dual-license complexity. > > I asked the scikit-learn folks as well, and they're saying something > varying between "fair use" and "in the spirit of open source" , which > I think comes down to "we don't really care". Of course, I don't care > either other than wanting to make sure people feel very free to use, > modify and show our examples, without caring too much about the > license. They'll probably do that anyway...

2 1

Re: Examples relicensing proposal
by Stéfan van der Walt 28 Jan '14

28 Jan '14

On Tue, Jan 28, 2014 at 1:42 PM, Juan Nunez-Iglesias <jni.soma(a)gmail.com> wrote: > Having read CC0, I'd be ok with the examples being under that license. I'll > let you decide whether you prefer the dual-license complexity. I asked the scikit-learn folks as well, and they're saying something varying between "fair use" and "in the spirit of open source" , which I think comes down to "we don't really care". Of course, I don't care either other than wanting to make sure people feel very free to use, modify and show our examples, without caring too much about the license. They'll probably do that anyway... Stéfan

1 0