[Insight-users] Canny Edge Detection Filter Qns., etc.

[Luis Ibanez]

Hi Kalpathi,

1. The thresholds in the Canny Edge Detection filter implemented in
    ITK apply to the product of the Second Derivative and the image
    Gradient.

    Please see the code in

            Insight/Code/BasicFilters/
               itkCannyEdgeDetectionImageFilter.txx

     Lines 370-407


::HysteresisThresholding()
{
// This is the Zero crossings of the Second derivative
// multiplied with the gradients of the image.
// HysteresisThresholding of this image should give
// the Canny output.
   typename OutputImageType::Pointer input =
                   m_MultiplyImageFilter->GetOutput();





2. The .mhd file header is part of the MetaImage file format.
    For details about this format, please refer to the
    ITK Software Guide:

        http://www.itk.org/ItkSoftwareGuide.pdf

    and the Data page of www.itk.org

        http://www.itk.org/HTML/Data.htm

    where you will find instructions on how to create .mhd files
    for your raw images.


    Also the FAQ:

http://www.itk.org/Wiki/ITK_FAQ#What_3D_file_formats_can_ITK_import_and_export.3F







3.  I couldn't find the colon datasets that you are refering to
     in your email at nova.nlm.nih.gov.

     I presume that they are CT datasets..., is that right ?


     You should try using the Example:

              Insight/Examples/IO
                  DicomSeriesReadImageWrite2.cxx


     It will convert a Dicom series into another file format
     such as Analyze, MetaImage, or GIPL.


     What DICOM viewer were you using ?





Regards,



    Luis



-------------------------
Kalpathi Subramanian wrote:
> 1. I am wondering about the two hysteresis thresholds (lower and
> upper thresholds) in this filter.
> If I remember this (I implemented all of this in my post-doc BSP tree
> project many, many years back), the thresholds were on the 
> CDF(cumulative distr. fn) of the gradient magnitudes. Is this the case 
> here? Or are the thresholds on
> just the gradient magnitude (maybe my implementation was based on
> another related implementation, I dont quite remember). Anyhow, this
> had the advantage that I could give these thresholds as percentages
> on the CDF..
> 
> 2.   - on reading raw data. There is a .mhd file needed
> to read raw data. Can someone point me to its format?
> 
> 3. Has anyone read the Colon datasets available from NLM 
> (nova.nlm.nih.gov) - I have tried it with the itkDicomSeriesReader
> but have had problems.. will try again but the images didnt
> seem like the original CT data; instead looked like binary images..
> 
> 
> Thanks.
> 
>     -- krs
> 
>