[Insight-users] Question about edge extraction

[Zachary Pincus]

Jiang,

As you will note from the tutorials online about the canny filtering=20
(and I *highly* recommend reading up on any method you intend to use;=20
everything just makes more sense that way!), the most important=20
parameter is the threshold for labeling edges that the canny filter=20
uses. (this is set with the SetThreshold method.)

This threshold controls how strong an edge must be before the filter=20
will attempt to trace it. Try setting that threshold; the default value=20=

is 0 (I think), which basically disables one of the most useful parts=20
of the canny filter.

The variance and error parameters (as per the ITK documentation)=20
control the initial smoothing, and should matter some, but a lot less=20
than the canny threshold.

Zach

On May 4, 2004, at 2:33 AM, Jiang wrote:

> Hi Zach,
> I figure the problem of using CannyEdgeDetection out. Since I render=20=

> the
> output of the CannyEdgeDetection directly. When I rescale the output =
to
> [0,255], then the result is correct.
> However there are a lot of noises in the output image. I consider it
> because of
> CannyFilter->SetVariance(1.0f);
> CannyFilter->SetMaximumError(.01f);
> How can I set an appropriate parameters for them?
> Thanks a lot!
>
>
> Chunyan Jiang
>
> -----Urspr=FCngliche Nachricht-----
> Von: Zachary Pincus [mailto:zpincus@stanford.edu]
> Gesendet: Dienstag, 4. Mai 2004 09:44
> An: Jiang
> Cc: ITK
> Betreff: Re: [SPAM:#] [Insight-users] Question about edge extraction
>
> The Canny edge detection filter may prove useful here. (I suggest
> looking over the myriad references about Canny edge detection on the
> web, such as: http://homepages.inf.ed.ac.uk/rbf/HIPR2/canny.htm -- the
> entire tutorial at this site is good.) A zero crossing filter may also
> be good (tutorial also at the above site).
>
> A canny filter (and to a lesser extent the zero crossing filter) is
> nice because the edges are constrained to be one pixel wide, and can
> extend through regions of low signal.
>
> Zach
>
>
> On May 4, 2004, at 12:22 AM, Jiang wrote:
>
>> Dear ITK users,
>>
>> I want to extract the edge picture from one image. And I use
>>
>>  itk::GradientMagnitudeImageFilter
>>
>> It works fine. However the output is float type. I want to get binary
>> output image. If I use
>>
>>  itk::BinaryThresholdImageFilter
>>
>> to filter the output to binary, and simply set
>> SetUpperThreshold(),SetLowerThreshold() to some defined value, the
>> binary edge is not good,
>>
>>  either with too much noise or the edge is not complete.
>>
>>  Could some one tell me how can I get good binary edge extraction
>> result?
>>
>> =A0
>>
>> Thanks a lot!
>>
>> =A0
>>
>> =A0
>>
>> Chunyan Jiang
>
> _______________________________________________
> Insight-users mailing list
> Insight-users@itk.org
> http://www.itk.org/mailman/listinfo/insight-users
>