[Insight-users] Edge Linking Problem

[Zachary Pincus]

Perhaps you should more carefully review the Software Guide, which=20
clearly explains what the advection (and other) terms in the level set=20=

methods refer to. (See esp. the section on Geodesic Active Contour=20
filters.)

Additionally, the Software Guide explains about the matter of seed=20
points for the GAC Filter -- instead of seed points that filter takes=20
an initial level set image as input. Please look at the copious=20
examples and documentation in the Guide, and the Doxygen documentation=20=

( http://www.itk.org/Doxygen/html/ ).

Additionally, I cannot "please explain to me why CannyEdgeDetection=20
Filter fails to detect the edges of an object" since you have provided=20=

so little information about the image you are using, or the parameters=20=

you have given to the filter. Perhaps you should try a range of=20
different parameters and review the copious information online about=20
how John Canny's edge detection method works.Knowing how a method works=20=

is in general very important for trying to use that method properly.=20
Try this: http://www.cee.hw.ac.uk/hipr/html/canny.html




On Mar 22, 2004, at 6:55 AM, CSPL wrote:

> Hello Zach Pincus,
> =A0
> =A0Thank you for that suggestion of yours, but I could not implement=20=

> your suggestion.
> =A0Firstly, I have already written the 8-connected code for the=20
> ConnectedComponentImage filter. For most of my needs just using the=20
> 8-connected filter was sufficient, but surely there will be some case=20=

> where this will fail.
> =A0
> =A0I tried to use the Geodesic Active Contour Filter, but could not =
get=20
> any result out of it. I could not find how to set the seed point for=20=

> this filter. Even the example in ITK Software Guide does not use seed=20=

> points.
>  =A0
> =A0I could not understand what advection term and the components of =
the=20
> equation are.
> Can you please explain to me why CannyEdgeDetection Filter fails to=20
> detect the edges of an object. For the same object, the gardient=20
> magnitude image filter is giving me better results.
>  =A0
> =A0Some times the canny filter is unable to identify even 25% of the=20=

> entire trace (i.e., not even the top hemisphere of an elliptical=20
> object). There is a huge difference in intensities of the required=20
> object and its immediate background. It looks as if a white object is=20=

> placed on a black background.
>  =A0
> =A0Can I in some way use this information of "Why Canny fails?" rather=20=

> than trying to improve its result by adding more filters to the=20
> application.
> =A0
> Regards,
> Rama Krishna.