[vtkusers] VtkConnectivity filter

[Richard Frank]

Thanks David,

Yes, I looked at the threshold connectivity filter and noticed the required seeds.

I have some 2d connected components/union find code that I will try to reimplement  for 3d.

I just wanted to see if the connectivity filter would do what I want, and it does seem to....

Do you happen to know if ITK has a general cc region find in n dimensions? We may want to link our app at some point to make use of ITK ( in spite of the different image layouts).

I think your class would be a good contribution VTK- it's a basic tool that seems to be missing....

Thanks

Rick Frank

On Mar 12, 2013, at 7:57 AM, David Gobbi <david.gobbi at gmail.com> wrote:

> Hi Richard,
> 
> If your data is a volume, then vtkImageThresholdConnectivity might be
> more suitable.  It requires you to set seeds, though, and I'm not sure
> if that's what you want.
> 
> I also have a filter in my personal collection that selects image
> regions based on size (rather than requiring you to set seed
> locations) and I could potentially contribute it to VTK.  Internally
> it computes labels for each region, up to a max of 254, and the code
> could be modified so that it outputs the labels.
> 
> - David
> 
> On Tue, Mar 12, 2013 at 5:43 AM, Richard Frank <rickfrank at me.com> wrote:
>> 
>> I'm having trouble understanding the usages of the connectivity filter.
>> 
>> I'm trying to use this filter to find and filter out regions in a volume, as one might do with bwlabeln() in matlab....
>> 
>> What does AddRegion() do? Does that usage imply that I already know how many regions there are? Why/when would I use DeleteRegion()?
>> 
>> 
>> If I ask for all regions, I get the whole volume even if I set a tight scalar threshold. If I ask for the largest region, I get what I would expect,  but that's not what I need.
>> 
>> If I set specific regions, and add 0,1,2 and use a threshold that I know should work, I get nothing.
>> 
>> So, I guess I don't understand the correct use.
>> 
>> Any advice welcome.....
>> 
>> 
>> Thanks
>> Rick Frank