Mailman 3 December 2011 - scikit-image

Re: watershed issue
by jip Dec. 31, 2011

Dec. 31, 2011

Thank you Tony, Modifying the threshold value from 30 to a lower value, gives a consistent result. Jean-Patrick sk.morphology.watershed(bassin_hi,markers,mask=top>30) On 30 déc, 17:22, Tony Yu <tsy...(a)gmail.com> wrote: > On Fri, Dec 30, 2011 at 8:01 AM, Tony Yu <tsy...(a)gmail.com> wrote: > > > On Fri, Dec 30, 2011 at 5:39 AM, jip <jeanpatrick.pomm...(a)gmail.com>wrote: > > >> Dear Emmanuelle, > > >> Sorry for the delay. > >> I have rewritten a clearer script > >> <https://docs.google.com/open?id=0B-YDFMbEy1grODU1Y2U0OTgtMzk1MS00Mzkx...>I > >> hope. > >> First, it takes a 12bits image<https://docs.google.com/open?id=0B-YDFMbEy1grNWJlYTkwYzYtMjk0ZS00NGVl...>(cy3 in the archive), isolates a region andremoves the background<https://docs.google.com/open?id=0B-YDFMbEy1grNDZjOWZjZjAtMGVmMS00ZTM2...> > >> . > >> Regional maxima are found to get markers for watershed. > >> Three kind of maps are also prepared. > >> Each map, with the markers image, is submitted to the watershed > >> algorithm.Segmentation is performed with watershed as indicated here<http://scikits-image.org/docs/dev/auto_examples/plot_watershed.html#e...>, > >> as follow: > >> segmentation=watershed(map,markers,mask) > >> map:grayscale image > >> markers:label image > >> mask:binary image > > > There's a small typo here: watershed should be called as: > > > watershed(map, makers, mask=mask) > > > More specifically, in your code, you have: > > > sk.morphology.watershed(bassin_hi,markers,top>30) > > > but it should be called as: > > > sk.morphology.watershed(bassin_hi,markers,mask=top>30) > > > Without specifying `mask=`, you're setting the connectivity parameter in > > watershed. Even with this correction, something seems to be off since > > watershed seems to return blank images. Without the masks, it gives what I > > would expect. I'm not sure what's going on the masked output. > > > -Tony > > Actually, I think this masked output is correct: All the markers seem to > land in the masked-out regions. > > >>> mask = top > 30 > >>> print np.any(marker * mask) > > False > > -Tony

1 0

Re: Need an advice
by Tony Yu Dec. 31, 2011

Dec. 31, 2011

On Fri, Dec 30, 2011 at 5:48 PM, Ralf Gommers <ralf.gommers(a)googlemail.com>wrote: > > > On Fri, Dec 30, 2011 at 8:05 PM, Tony Yu <tsyu80(a)gmail.com> wrote: > >> >> >> On Fri, Dec 30, 2011 at 1:19 PM, Ralf Gommers < >> ralf.gommers(a)googlemail.com> wrote: >> >>> >>> >>> On Fri, Dec 30, 2011 at 5:46 PM, Tony Yu <tsyu80(a)gmail.com> wrote: >>> >>>> >>>> >>>> On Fri, Dec 30, 2011 at 8:20 AM, Nadav Horesh <nadavh.horesh(a)gmail.com>wrote: >>>> >>>>> Remarks for the convolution subpackage: >>>>> 1. The __init__.py should be probably changed to >>>>> >>>>> from ext import pyconvolve as convolve >>>>> >>>> >>>> Agreed. Alternatively, the c-method could be renamed so that the python >>>> method in ext can be named convolve. >>>> >>>>> >>>>> 2. The test_convolution.py file is very instructive. It should be >>>>> polished a bit. I think it would be worthwhile to add also >>>>> scipy.signal.sepfir2d and cv2.sepFilter2D to the test. >>>>> >>>> >>>>> 3. I found the convolve speed in par with cv.convolve2d (it can be >>>>> slower or faster depending on the kernel) >>>>> >>>>> Nadav >>>>> >>>> >>>> Thanks for testing the code. >>>> >>>> I've got the code compiling on my system now, so I was able to clean up >>>> a few things. Could you update to what's currently on my >>>> skimage-convolution branch to check that it builds on your system? If not, >>>> what flags do you have to change/add to get it to compile? >>>> >>>> -Tony >>>> >>>> P.S. If any one else is building this on OS X: ppc compatible builds is >>>> turned on by default (at least on my system). This should be turned off >>>> before building. Add the following line to ~/.profile, or just run it right >>>> before building: >>>> >>>> export ARCH_FLAGS="-arch i386 -arch x86_64" >>>> >>> >>> Can you explain why is this necessary? >>> >>> In convolution/setup.py I also see you add "-ffast-math", which is a >>> GCC-specific flag. It's better to remove it I think, it may give problems >>> with other compilers. >>> >>> Ralf >>> >> >> On my system at least, running gcc will, by default, add "-arch ppc" >> along with i386 and x86_64. As far as I can tell, PPC doesn't support SSE, >> which is used by this implementation. >> > > That's true. I can't tell for sure, but it seems that this code is not > optional. Meaning PPC is unsupported, and the ARCH_FLAGS need to be defined > on OS X. I'm only a very occasional user now so it doesn't bother me too > much, but isn't this a large penalty for the benefit you get? > Well, I assume that we would be able to add a check to disable the build on unsupported architectures. As far as, whether it's worth it: I don't know. Is PPC very common? I'm only familiar with it in the context of old Apple computers. From the Wikipedia page <http://en.wikipedia.org/wiki/PowerPC>, it seems like PPC is mainly used in servers now. As for the benefit: I think getting close to the performance of OpenCV is pretty significant. (On my system, however, I don't quite see the performance boost that Nadav sees: skimage: 0.034 secs, ndimage: 0.089 secs, opencv: 0.018 secs, for the output of 'test_convolution.py'---nevertheless, it's a significant improvement over ndimage). And, in comparison, OpenCV can be incredibly difficult to build. There are probably other ways to get similar performance, but like I said, I don't do any serious non-Python coding, so I have no idea what those alternatives may be. Stefan and Pieter (the original author of this code) have probably thought this through a bit more. -Tony >> The flag is left over from the original implementation. Removing >> "-ffast-math" works fine on my system. To be honest, I don't do much >> non-Python coding so I don't know the implications of adding/removing the >> flag. >> > > It's very likely to give problems with unusual (or even Intel) compilers I > think. > > Ralf > >

1 0

Re: Need an advice
by Ralf Gommers Dec. 31, 2011

Dec. 31, 2011

On Fri, Dec 30, 2011 at 8:05 PM, Tony Yu <tsyu80(a)gmail.com> wrote: > > > On Fri, Dec 30, 2011 at 1:19 PM, Ralf Gommers <ralf.gommers(a)googlemail.com > > wrote: > >> >> >> On Fri, Dec 30, 2011 at 5:46 PM, Tony Yu <tsyu80(a)gmail.com> wrote: >> >>> >>> >>> On Fri, Dec 30, 2011 at 8:20 AM, Nadav Horesh <nadavh.horesh(a)gmail.com>wrote: >>> >>>> Remarks for the convolution subpackage: >>>> 1. The __init__.py should be probably changed to >>>> >>>> from ext import pyconvolve as convolve >>>> >>> >>> Agreed. Alternatively, the c-method could be renamed so that the python >>> method in ext can be named convolve. >>> >>>> >>>> 2. The test_convolution.py file is very instructive. It should be >>>> polished a bit. I think it would be worthwhile to add also >>>> scipy.signal.sepfir2d and cv2.sepFilter2D to the test. >>>> >>> >>>> 3. I found the convolve speed in par with cv.convolve2d (it can be >>>> slower or faster depending on the kernel) >>>> >>>> Nadav >>>> >>> >>> Thanks for testing the code. >>> >>> I've got the code compiling on my system now, so I was able to clean up >>> a few things. Could you update to what's currently on my >>> skimage-convolution branch to check that it builds on your system? If not, >>> what flags do you have to change/add to get it to compile? >>> >>> -Tony >>> >>> P.S. If any one else is building this on OS X: ppc compatible builds is >>> turned on by default (at least on my system). This should be turned off >>> before building. Add the following line to ~/.profile, or just run it right >>> before building: >>> >>> export ARCH_FLAGS="-arch i386 -arch x86_64" >>> >> >> Can you explain why is this necessary? >> >> In convolution/setup.py I also see you add "-ffast-math", which is a >> GCC-specific flag. It's better to remove it I think, it may give problems >> with other compilers. >> >> Ralf >> > > On my system at least, running gcc will, by default, add "-arch ppc" along > with i386 and x86_64. As far as I can tell, PPC doesn't support SSE, which > is used by this implementation. > That's true. I can't tell for sure, but it seems that this code is not optional. Meaning PPC is unsupported, and the ARCH_FLAGS need to be defined on OS X. I'm only a very occasional user now so it doesn't bother me too much, but isn't this a large penalty for the benefit you get? > > The flag is left over from the original implementation. Removing > "-ffast-math" works fine on my system. To be honest, I don't do much > non-Python coding so I don't know the implications of adding/removing the > flag. > It's very likely to give problems with unusual (or even Intel) compilers I think. Ralf

1 0

Re: Need an advice
by Tony Yu Dec. 31, 2011

Dec. 31, 2011

On Fri, Dec 30, 2011 at 1:19 PM, Ralf Gommers <ralf.gommers(a)googlemail.com>wrote: > > > On Fri, Dec 30, 2011 at 5:46 PM, Tony Yu <tsyu80(a)gmail.com> wrote: > >> >> >> On Fri, Dec 30, 2011 at 8:20 AM, Nadav Horesh <nadavh.horesh(a)gmail.com>wrote: >> >>> Remarks for the convolution subpackage: >>> 1. The __init__.py should be probably changed to >>> >>> from ext import pyconvolve as convolve >>> >> >> Agreed. Alternatively, the c-method could be renamed so that the python >> method in ext can be named convolve. >> >>> >>> 2. The test_convolution.py file is very instructive. It should be >>> polished a bit. I think it would be worthwhile to add also >>> scipy.signal.sepfir2d and cv2.sepFilter2D to the test. >>> >> >>> 3. I found the convolve speed in par with cv.convolve2d (it can be >>> slower or faster depending on the kernel) >>> >>> Nadav >>> >> >> Thanks for testing the code. >> >> I've got the code compiling on my system now, so I was able to clean up a >> few things. Could you update to what's currently on my skimage-convolution >> branch to check that it builds on your system? If not, what flags do you >> have to change/add to get it to compile? >> >> -Tony >> >> P.S. If any one else is building this on OS X: ppc compatible builds is >> turned on by default (at least on my system). This should be turned off >> before building. Add the following line to ~/.profile, or just run it right >> before building: >> >> export ARCH_FLAGS="-arch i386 -arch x86_64" >> > > Can you explain why is this necessary? > > In convolution/setup.py I also see you add "-ffast-math", which is a > GCC-specific flag. It's better to remove it I think, it may give problems > with other compilers. > > Ralf > On my system at least, running gcc will, by default, add "-arch ppc" along with i386 and x86_64. As far as I can tell, PPC doesn't support SSE, which is used by this implementation. The flag is left over from the original implementation. Removing "-ffast-math" works fine on my system. To be honest, I don't do much non-Python coding so I don't know the implications of adding/removing the flag. -Tony

1 0

Re: Need an advice
by Ralf Gommers Dec. 31, 2011

Dec. 31, 2011

On Fri, Dec 30, 2011 at 5:46 PM, Tony Yu <tsyu80(a)gmail.com> wrote: > > > On Fri, Dec 30, 2011 at 8:20 AM, Nadav Horesh <nadavh.horesh(a)gmail.com>wrote: > >> Remarks for the convolution subpackage: >> 1. The __init__.py should be probably changed to >> >> from ext import pyconvolve as convolve >> > > Agreed. Alternatively, the c-method could be renamed so that the python > method in ext can be named convolve. > >> >> 2. The test_convolution.py file is very instructive. It should be >> polished a bit. I think it would be worthwhile to add also >> scipy.signal.sepfir2d and cv2.sepFilter2D to the test. >> > >> 3. I found the convolve speed in par with cv.convolve2d (it can be slower >> or faster depending on the kernel) >> >> Nadav >> > > Thanks for testing the code. > > I've got the code compiling on my system now, so I was able to clean up a > few things. Could you update to what's currently on my skimage-convolution > branch to check that it builds on your system? If not, what flags do you > have to change/add to get it to compile? > > -Tony > > P.S. If any one else is building this on OS X: ppc compatible builds is > turned on by default (at least on my system). This should be turned off > before building. Add the following line to ~/.profile, or just run it right > before building: > > export ARCH_FLAGS="-arch i386 -arch x86_64" > Can you explain why is this necessary? In convolution/setup.py I also see you add "-ffast-math", which is a GCC-specific flag. It's better to remove it I think, it may give problems with other compilers. Ralf > > > >> >> 2011/12/28 Nadav Horesh <nadavh.horesh(a)gmail.com> >> >>> I compiled after setting the flags for a core2 linux machine. A quick >>> and dirty benchmark showed a x2 improvement over scipy.signal.sepfir2d. >>> I think that are some declaration errors in the python code (for >>> instance I think that the convolution __init__.py should be >>> >>> from ext import pyconvolve >>> >>> instead of >>> >>> from ext import convolve >>> >>> I'll take a closer look later. >>> >>> Thank you, >>> >>> Nadav. >>> >>> >>> >>> 2011/12/28 Tony Yu <tsyu80(a)gmail.com> >>> >>>> >>>> >>>> On Tue, Dec 27, 2011 at 8:24 PM, Tony Yu <tsyu80(a)gmail.com> wrote: >>>> >>>>> >>>>> >>>>> On Tue, Dec 27, 2011 at 8:29 AM, Nadav Horesh <nadavh.horesh(a)gmail.com >>>>> > wrote: >>>>> >>>>>> How can I get it? >>>>>> >>>>>> Nadav >>>>>> >>>>> >>>>> The code Stefan mentioned is in an old pull request (PR #16<https://github.com/scikits-image/scikits-image/pull/16>). >>>>> That PR is from before the switch from scikits.image to skimage. I've made >>>>> a new branch<https://github.com/tonysyu/scikits-image/commits/skimage-convolution>in my git repo with the convolution code in the new namespace. >>>>> >>>>> If you're already running skimage from git, then you can use git to >>>>> clone this branch: >>>>> >>>>> git remote -f add tonysyu git@github.com:tonysyu/scikits-image.git >>>>> git checkout -b convolution tonysyu/skimage-convolution >>>>> >>>>> >>>>> In case this isn't familiar: The first line adds my git repo to your >>>>> list of remotes, and the "-f" flag fetches the tags from my repo (this >>>>> ensures that you have the info about my branches). My repo is now added to >>>>> your repo as a remote with the name "tonysyu" (you can change this). The >>>>> second line creates a new branch "convolution", clones my >>>>> "skimage-convolution" branch into it, and checks it out. >>>>> >>>>> Hope that helps, >>>>> -Tony >>>>> >>>> >>>> Oops, I forgot to mention: I added a compile flag that is specific to >>>> my system (In convolution/setup.py there's an include flag '-I/usr/...') >>>> that should probably be removed. Also, I should mention that I haven't >>>> gotten this branch to compile on my system (OSX). I believe it should >>>> compile OK on Linux (after removing the '-I/usr/...' compile flag). >>>> >>>> Good luck, >>>> -Tony >>>> >>>>> >>>>> 2011/12/23 Stéfan van der Walt <stefan(a)sun.ac.za> >>>>> >>>>>> On Thu, Dec 22, 2011 at 10:44 PM, Nadav Horesh < >>>>>> nadavh.horesh(a)gmail.com> wrote: >>>>>> > The application is not too sophisticated, and I do use numpy, scipy >>>>>> and >>>>>> > skimage for all I need. The only problem is speed, I need the >>>>>> processing to >>>>>> > be at l5-10 time faster. Convolutions take most of the processing >>>>>> time. I >>>>>> > prefer a flexible solution that would speed up also common >>>>>> nonlinear filters >>>>>> > (i.e. a median filter). >>>>>> >>>>>> We have a PR in the pipeline for doing really, really fast >>>>>> convolutions... but it needs a bit of work. Would you like to have a >>>>>> look at it? >>>>>> >>>>>> Stéfan >>>>>> >>>>> >>>> >>>> >

1 0

Re: Need an advice
by Stéfan van der Walt Dec. 31, 2011

Dec. 31, 2011

On Fri, Dec 30, 2011 at 8:20 AM, Nadav Horesh <nadavh.horesh(a)gmail.com> wrote: > Remarks for the convolution subpackage: > 1. The __init__.py should be probably changed to > > from ext import pyconvolve as convolve > > 2. The test_convolution.py file is very instructive. It should be polished a > bit. I think it would be worthwhile to add also scipy.signal.sepfir2d > and cv2.sepFilter2D to the test. > > 3. I found the convolve speed in par with cv.convolve2d (it can be slower or > faster depending on the kernel) That's great. Do you know how we can reliably make this code compile on all supported platforms? Stéfan

1 0

Re: Need an advice
by Tony Yu Dec. 30, 2011

Dec. 30, 2011

On Fri, Dec 30, 2011 at 8:20 AM, Nadav Horesh <nadavh.horesh(a)gmail.com>wrote: > Remarks for the convolution subpackage: > 1. The __init__.py should be probably changed to > > from ext import pyconvolve as convolve > Agreed. Alternatively, the c-method could be renamed so that the python method in ext can be named convolve. > > 2. The test_convolution.py file is very instructive. It should be polished > a bit. I think it would be worthwhile to add also scipy.signal.sepfir2d > and cv2.sepFilter2D to the test. > > 3. I found the convolve speed in par with cv.convolve2d (it can be slower > or faster depending on the kernel) > > Nadav > Thanks for testing the code. I've got the code compiling on my system now, so I was able to clean up a few things. Could you update to what's currently on my skimage-convolution branch to check that it builds on your system? If not, what flags do you have to change/add to get it to compile? -Tony P.S. If any one else is building this on OS X: ppc compatible builds is turned on by default (at least on my system). This should be turned off before building. Add the following line to ~/.profile, or just run it right before building: export ARCH_FLAGS="-arch i386 -arch x86_64" > > 2011/12/28 Nadav Horesh <nadavh.horesh(a)gmail.com> > >> I compiled after setting the flags for a core2 linux machine. A quick >> and dirty benchmark showed a x2 improvement over scipy.signal.sepfir2d. >> I think that are some declaration errors in the python code (for instance >> I think that the convolution __init__.py should be >> >> from ext import pyconvolve >> >> instead of >> >> from ext import convolve >> >> I'll take a closer look later. >> >> Thank you, >> >> Nadav. >> >> >> >> 2011/12/28 Tony Yu <tsyu80(a)gmail.com> >> >>> >>> >>> On Tue, Dec 27, 2011 at 8:24 PM, Tony Yu <tsyu80(a)gmail.com> wrote: >>> >>>> >>>> >>>> On Tue, Dec 27, 2011 at 8:29 AM, Nadav Horesh <nadavh.horesh(a)gmail.com>wrote: >>>> >>>>> How can I get it? >>>>> >>>>> Nadav >>>>> >>>> >>>> The code Stefan mentioned is in an old pull request (PR #16<https://github.com/scikits-image/scikits-image/pull/16>). >>>> That PR is from before the switch from scikits.image to skimage. I've made >>>> a new branch<https://github.com/tonysyu/scikits-image/commits/skimage-convolution>in my git repo with the convolution code in the new namespace. >>>> >>>> If you're already running skimage from git, then you can use git to >>>> clone this branch: >>>> >>>> git remote -f add tonysyu git@github.com:tonysyu/scikits-image.git >>>> git checkout -b convolution tonysyu/skimage-convolution >>>> >>>> >>>> In case this isn't familiar: The first line adds my git repo to your >>>> list of remotes, and the "-f" flag fetches the tags from my repo (this >>>> ensures that you have the info about my branches). My repo is now added to >>>> your repo as a remote with the name "tonysyu" (you can change this). The >>>> second line creates a new branch "convolution", clones my >>>> "skimage-convolution" branch into it, and checks it out. >>>> >>>> Hope that helps, >>>> -Tony >>>> >>> >>> Oops, I forgot to mention: I added a compile flag that is specific to my >>> system (In convolution/setup.py there's an include flag '-I/usr/...') that >>> should probably be removed. Also, I should mention that I haven't gotten >>> this branch to compile on my system (OSX). I believe it should compile OK >>> on Linux (after removing the '-I/usr/...' compile flag). >>> >>> Good luck, >>> -Tony >>> >>>> >>>> 2011/12/23 Stéfan van der Walt <stefan(a)sun.ac.za> >>>> >>>>> On Thu, Dec 22, 2011 at 10:44 PM, Nadav Horesh < >>>>> nadavh.horesh(a)gmail.com> wrote: >>>>> > The application is not too sophisticated, and I do use numpy, scipy >>>>> and >>>>> > skimage for all I need. The only problem is speed, I need the >>>>> processing to >>>>> > be at l5-10 time faster. Convolutions take most of the processing >>>>> time. I >>>>> > prefer a flexible solution that would speed up also common nonlinear >>>>> filters >>>>> > (i.e. a median filter). >>>>> >>>>> We have a PR in the pipeline for doing really, really fast >>>>> convolutions... but it needs a bit of work. Would you like to have a >>>>> look at it? >>>>> >>>>> Stéfan >>>>> >>>> >>> >>>

1 0

Re: watershed issue
by Tony Yu Dec. 30, 2011

Dec. 30, 2011

On Fri, Dec 30, 2011 at 8:01 AM, Tony Yu <tsyu80(a)gmail.com> wrote: > > > On Fri, Dec 30, 2011 at 5:39 AM, jip <jeanpatrick.pommier(a)gmail.com>wrote: > >> Dear Emmanuelle, >> >> Sorry for the delay. >> I have rewritten a clearer script >> <https://docs.google.com/open?id=0B-YDFMbEy1grODU1Y2U0OTgtMzk1MS00MzkxLTliOW…>I >> hope. >> First, it takes a 12bits image<https://docs.google.com/open?id=0B-YDFMbEy1grNWJlYTkwYzYtMjk0ZS00NGVlLWFjMz…>(cy3 in the archive), isolates a region andremoves the background<https://docs.google.com/open?id=0B-YDFMbEy1grNDZjOWZjZjAtMGVmMS00ZTM2LTgwYz…> >> . >> Regional maxima are found to get markers for watershed. >> Three kind of maps are also prepared. >> Each map, with the markers image, is submitted to the watershed >> algorithm.Segmentation is performed with watershed as indicated here<http://scikits-image.org/docs/dev/auto_examples/plot_watershed.html#example…>, >> as follow: >> segmentation=watershed(map,markers,mask) >> map:grayscale image >> markers:label image >> mask:binary image >> >> > There's a small typo here: watershed should be called as: > > watershed(map, makers, mask=mask) > > More specifically, in your code, you have: > > sk.morphology.watershed(bassin_hi,markers,top>30) > > but it should be called as: > > sk.morphology.watershed(bassin_hi,markers,mask=top>30) > > Without specifying `mask=`, you're setting the connectivity parameter in > watershed. Even with this correction, something seems to be off since > watershed seems to return blank images. Without the masks, it gives what I > would expect. I'm not sure what's going on the masked output. > > -Tony > > Actually, I think this masked output is correct: All the markers seem to land in the masked-out regions. >>> mask = top > 30 >>> print np.any(marker * mask) False -Tony

1 0

Re: Need an advice
by Nadav Horesh Dec. 30, 2011

Dec. 30, 2011

Remarks for the convolution subpackage: 1. The __init__.py should be probably changed to from ext import pyconvolve as convolve 2. The test_convolution.py file is very instructive. It should be polished a bit. I think it would be worthwhile to add also scipy.signal.sepfir2d and cv2.sepFilter2D to the test. 3. I found the convolve speed in par with cv.convolve2d (it can be slower or faster depending on the kernel) Nadav 2011/12/28 Nadav Horesh <nadavh.horesh(a)gmail.com> > I compiled after setting the flags for a core2 linux machine. A quick and > dirty benchmark showed a x2 improvement over scipy.signal.sepfir2d. > I think that are some declaration errors in the python code (for instance > I think that the convolution __init__.py should be > > from ext import pyconvolve > > instead of > > from ext import convolve > > I'll take a closer look later. > > Thank you, > > Nadav. > > > > 2011/12/28 Tony Yu <tsyu80(a)gmail.com> > >> >> >> On Tue, Dec 27, 2011 at 8:24 PM, Tony Yu <tsyu80(a)gmail.com> wrote: >> >>> >>> >>> On Tue, Dec 27, 2011 at 8:29 AM, Nadav Horesh <nadavh.horesh(a)gmail.com>wrote: >>> >>>> How can I get it? >>>> >>>> Nadav >>>> >>> >>> The code Stefan mentioned is in an old pull request (PR #16<https://github.com/scikits-image/scikits-image/pull/16>). >>> That PR is from before the switch from scikits.image to skimage. I've made >>> a new branch<https://github.com/tonysyu/scikits-image/commits/skimage-convolution>in my git repo with the convolution code in the new namespace. >>> >>> If you're already running skimage from git, then you can use git to >>> clone this branch: >>> >>> git remote -f add tonysyu git@github.com:tonysyu/scikits-image.git >>> git checkout -b convolution tonysyu/skimage-convolution >>> >>> >>> In case this isn't familiar: The first line adds my git repo to your >>> list of remotes, and the "-f" flag fetches the tags from my repo (this >>> ensures that you have the info about my branches). My repo is now added to >>> your repo as a remote with the name "tonysyu" (you can change this). The >>> second line creates a new branch "convolution", clones my >>> "skimage-convolution" branch into it, and checks it out. >>> >>> Hope that helps, >>> -Tony >>> >> >> Oops, I forgot to mention: I added a compile flag that is specific to my >> system (In convolution/setup.py there's an include flag '-I/usr/...') that >> should probably be removed. Also, I should mention that I haven't gotten >> this branch to compile on my system (OSX). I believe it should compile OK >> on Linux (after removing the '-I/usr/...' compile flag). >> >> Good luck, >> -Tony >> >>> >>> 2011/12/23 Stéfan van der Walt <stefan(a)sun.ac.za> >>> >>>> On Thu, Dec 22, 2011 at 10:44 PM, Nadav Horesh <nadavh.horesh(a)gmail.com> >>>> wrote: >>>> > The application is not too sophisticated, and I do use numpy, scipy >>>> and >>>> > skimage for all I need. The only problem is speed, I need the >>>> processing to >>>> > be at l5-10 time faster. Convolutions take most of the processing >>>> time. I >>>> > prefer a flexible solution that would speed up also common nonlinear >>>> filters >>>> > (i.e. a median filter). >>>> >>>> We have a PR in the pipeline for doing really, really fast >>>> convolutions... but it needs a bit of work. Would you like to have a >>>> look at it? >>>> >>>> Stéfan >>>> >>> >> >>

1 0

watershed issue
by jip Dec. 30, 2011

Dec. 30, 2011

Dear all, I try to use watershed implemenation in skimage, compared with scipy as follow: # -*- coding: utf-8 -*- """ Created on Thu Dec 22 13:45:46 2011 @author: Jean-Patrick Pommier Telomeres quantification at low resolution with skimage """ import scipy.ndimage as nd import pylab as plb import numpy as np import skimage as sk import skimage.morphology as mo import skimage.io as io import skimage.filter.edges as edges #import mahotas as mh #import pymorph def gray12_to_8(im): '''Converts a 12 bits image to 8 bits ''' i=0.062271062*im return np.uint8(i) def Hipass(im,size=9): '''returns a hipass filtered image with background >=0 with a gaussian filter of size size=9 by default''' blur=blur=nd.gaussian_filter(im,size) hi=im-1.0*blur-5 mask=np.copy(hi) hi[mask<0]=0 return hi def gray_dist(im): '''Build a pseudo distance map from a high pass filtered image suitable for watershed''' f=Hipass(im,9) dist=f.max()-f dist -= dist.min() dist = dist/float(dist.ptp()) * 255 dist = dist.astype(np.uint8) return dist #============================================================================== # images #============================================================================== path="/home/claire/Applications/ImagesTest/JPP50/16/" CC="DAPI/" PR="CY3/" im="1.TIF" #sk.io.use_plugin('gdal', 'imread') sk.io.use_plugin('freeimage', 'imread') sk.io.use_plugin('matplotlib', 'imshow') dapi=io.imread(path+CC+im) cy3=io.imread(path+PR+im) #============================================================================= # Extracting known nuclei #============================================================================== nuc=dapi[215:264,1151:1198] telo=cy3[215:264,1151:1198] telo=np.uint16(telo) #============================================================================== # Telomeres Segmentation with skimage & ndimage print telo.dtype se=mo.disk(7) top=mo.greyscale_white_top_hat(gray12_to_8(telo),se) #print top.min(),top.max(),top.dtype hi=np.uint8(Hipass(top,size=9)) #print hi.min(), hi.max(),hi.dtype dist=nd.distance_transform_bf(top>10) bassin_dist=np.uint16(dist.max()-dist) bassin_hi=hi.max()-hi remax=mo.is_local_maximum(bassin_hi,hi>6) markers,n=nd.label(remax) print n, 'spots detected' edgesmap=edges.sobel(hi) seg=sk.morphology.watershed(edgesmap,markers,top>30) segn=nd.watershed_ift(bassin_hi,markers) segn_mark_dist=nd.watershed_ift(bassin_dist,markers) #============================================================================== # Quantification #============================================================================== print top.shape,top.dtype #print np.shape(seg),np.dtype(seg) #spots_I=nd.measurements.sum(top,seg) #print spots_I #============================================================================== # #============================================================================== plb.figure(1) plb.subplot(341,frameon=False, xticks=[], yticks=[]) plb.title('raw 12bits') plb.imshow(telo) plb.subplot(342,frameon=False, xticks=[], yticks=[]) plb.title('top Hat') plb.imshow(top) plb.subplot(343,frameon=False, xticks=[], yticks=[]) plb.title('hiP') plb.imshow(hi) plb.subplot(344,frameon=False, xticks=[], yticks=[]) plb.title('regional max') plb.imshow(remax,interpolation=None) plb.subplot(345,frameon=False, xticks=[], yticks=[]) plb.title('markers') plb.imshow(markers,interpolation=None) plb.subplot(346,frameon=False, xticks=[], yticks=[]) plb.title('bassin=hi') plb.imshow(bassin_hi,interpolation=None) plb.subplot(347,frameon=False, xticks=[], yticks=[]) plb.title('bassin dist') plb.imshow(bassin_dist,interpolation=None) plb.subplot(348,frameon=False, xticks=[], yticks=[]) plb.title('hi+mark (nd)') plb.imshow(segn,interpolation=None) plb.subplot(349,frameon=False, xticks=[], yticks=[]) plb.title('sobel (sk)') plb.imshow(edgesmap,interpolation=None) plb.subplot(3,4,10,frameon=False, xticks=[], yticks=[]) plb.title('sob+mark (sk)') plb.imshow(seg,interpolation=None) plb.subplot(3,4,12,frameon=False, xticks=[], yticks=[]) plb.title('dist+mark (nd)') plb.imshow(segn_mark_dist,interpolation=None) plb.show() <https://lh5.googleusercontent.com/-OBPfUOGJNok/TvtKO_6q20I/AAAAAAAAAqw/6VX7…> I get a strange result, see the "sob+mark (sk)" image, compared to the segmentations obtained with the scipy implementation. Is it a skimage bug ? Best regards. Jean-Patrick

3 3