[ITK] Canny edge detection filter results

[Dženan Zukić]

Can you share the source code of the example which uses the Canny filter?

On Tue, Feb 2, 2016 at 4:24 PM, Joao Ascenso <j.ascenso at campus.fct.unl.pt>
wrote:

> That's what have done. Started a new project, new directory and just using
> the canny filter gives me that, but no problems without it.
> Forget to mention I'm working with MITK.
>
>
> On 2 February 2016 at 21:08, Dženan Zukić <dzenanz at gmail.com> wrote:
>
>> The safest cure for such things is to do a clean rebuild (create new
>> build directory, do cmake and then open .sln file with VS and build). Let
>> us know if you run into problems.
>>
>> HTH
>>
>> On Tue, Feb 2, 2016 at 4:01 PM, Joao Ascenso <j.ascenso at campus.fct.unl.pt
>> > wrote:
>>
>>> Hello again,
>>>
>>> For some reason this errors appear in itkSparceFieldLayer.hxx
>>>
>>> Error C2512 'itk::ListNode<itk::Index<3>>': no appropriate default
>>> constructor available
>>> Error C2027 use of undefined type 'itk::ListNode<itk::Index<3>>'
>>> Error C2227 left of '->Next' must point to class/struct/union/generic
>>> type
>>> Error C2027 use of undefined type 'itk::ListNode<itk::Index<3>>'
>>> Error C2227 left of '->Previous' must point to
>>> class/struct/union/generic type
>>>
>>> They are related with the compilation of
>>> itk::CannyEdgeDetectionImageFilter. I've been using it so far with no
>>> problem. It started when I run the cmake file again and it complained about
>>> Platform Toolset v140 not being installed. I changed back to v120 to run it
>>> and those errors started.
>>>
>>> Generated project again in cmake to make sure it was with Visual Studio
>>> 12, same result.
>>> Any clues on this? Thanks
>>>
>>> On 19 January 2016 at 17:05, Joao Ascenso <j.ascenso at campus.fct.unl.pt>
>>> wrote:
>>>
>>>> Thank you Dženan.
>>>>
>>>> I did some oversampling between slices to solve the problem, with the
>>>> downside of making it much slower.
>>>>
>>>>
>>>>
>>>> On 18 January 2016 at 21:26, Dženan Zukić <dzenanz at gmail.com> wrote:
>>>>
>>>>> Hi Joao,
>>>>>
>>>>> if you are doing Canny filtering in 3D and then looking at 2D slices,
>>>>> the edges are usually not 1 pixel thick. This is because edge for 3D is a
>>>>> surface, and when you slice that surface along anatomical axes the surface
>>>>> will not be orthogonal to the slices in most places. Where surface is
>>>>> orthogonal, you get nice 1 pixel wide intersections. Where surface is
>>>>> parallel to the slice, you get a sizable intersection surface.
>>>>>
>>>>> Regards,
>>>>> Dženan
>>>>>
>>>>> On Mon, Jan 18, 2016 at 11:47 AM, Joao Ascenso <
>>>>> j.ascenso at campus.fct.unl.pt> wrote:
>>>>>
>>>>>> Hello,
>>>>>> I'm working on OCT images (DICOM volume) of the retina and trying to
>>>>>> do the segmentation of some layers.
>>>>>>
>>>>>> After doing some noise reduction I'm applying the
>>>>>> CannyEdgeDetectionImageFilter, but the result give me thicker edges than
>>>>>> should be expected from Canny.
>>>>>>
>>>>>> Changing thresholds influences the amount of edges detected but don't
>>>>>> improve this issue.
>>>>>>
>>>>>> Anyone have came across similar problem? Any suggestion is welcome.
>>>>>>
>>>>>> Thank you,
>>>>>>
>>>>>> João Ascenso
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> Community mailing list
>>>>>> Community at itk.org
>>>>>> http://public.kitware.com/mailman/listinfo/community
>>>>>>
>>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> João Ascenso
>>>> Universidade Nova de Lisboa - Faculdade de Ciências e Tecnologias
>>>> Mestrado Integrado em Engenharia Electrotécnica e de Computadores
>>>>
>>>>
>>>>
>>>
>>>
>>> --
>>> João Ascenso
>>> Universidade Nova de Lisboa - Faculdade de Ciências e Tecnologias
>>> Mestrado Integrado em Engenharia Electrotécnica e de Computadores
>>>
>>>
>>>
>>
>
>
> --
> João Ascenso
> Universidade Nova de Lisboa - Faculdade de Ciências e Tecnologias
> Mestrado Integrado em Engenharia Electrotécnica e de Computadores
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/community/attachments/20160202/77ad7807/attachment-0001.html>