question about ultimate points

[Aryeh Weiss]

I have an image in which binary->ultimate points (also biary->watershed)
causes some objects to disappear. I have a small example but it cannot
be posted.   I think you can download it from
http://www.megaupload.com/?d=8SCDE952

If that does not work, then I will be happy to send anyone interested
the example (it is a binary image, 4KB zipped on my disk).

It may be related to the fact that all pixels in the disappearing object
are neighbors to a boundary. However, note that a point object is left
intact.

Is this a bug, or something that I dont know about the definition of
ultimate points? i expected it to leave something, but maybe certain
classes of objects have no well defined ultimate point. If so, I would
want to be able to find such objects so that I can deal with them
separately.

--aryeh
--
Aryeh Weiss
School of Engineering
Bar Ilan University
Ramat Gan 52900 Israel

Ph:  972-3-5317638
FAX: [phone]

[Gabriel Landini]

On Thursday 19 May 2011 11:53:23 you wrote:
> I have an image in which binary->ultimate points (also biary->watershed)
> causes some objects to disappear. I have a small example but it cannot
> be posted.   I think you can download it from
> http://www.megaupload.com/?d=8SCDE952


Yes, it looks like a bug to me

> It may be related to the fact that all pixels in the disappearing object
> are neighbors to a boundary. However, note that a point object is left
> intact.

Not necessarily, you can have an ordered erosion of the blob which has a
preferential  asymmetry which leads to a point even in the case of no-interior
blobs like the example.
If you add let's say a white disk somewhere then the blobs are detected! So I
think that the image is failing because just in the first iteration the
erosions get rid of the blob.

I hope this helps to narrow down the problem.

As an temporary measure you can skeletonise the image and then apply the
ulitimate points. That seems to work in your image.

Cheers
Gabriel

[Michael Schmid]

Hi Gabriel, Aryeh,

yes it is a bug in code contributed by me and I am working on how to  
solve it.
There is an easy fix, but I have to check whether it would slow down  
the code too much.

Michael
________________________________________________________________

On 19 May 2011, at 13:27, Gabriel Landini wrote:

> On Thursday 19 May 2011 11:53:23 you wrote:
>> I have an image in which binary->ultimate points (also biary-
>> >watershed)
>> causes some objects to disappear. I have a small example but it  
>> cannot
>> be posted.   I think you can download it from
>> http://www.megaupload.com/?d=8SCDE952
>
>
> Yes, it looks like a bug to me
>
>> It may be related to the fact that all pixels in the disappearing  
>> object
>> are neighbors to a boundary. However, note that a point object is  
>> left
>> intact.
>
> Not necessarily, you can have an ordered erosion of the blob which  
> has a
> preferential  asymmetry which leads to a point even in the case of  
> no-interior
> blobs like the example.
> If you add let's say a white disk somewhere then the blobs are  
> detected! So I
> think that the image is failing because just in the first iteration  
> the
> erosions get rid of the blob.
>
> I hope this helps to narrow down the problem.
>
> As an temporary measure you can skeletonise the image and then  
> apply the
> ulitimate points. That seems to work in your image.
>
> Cheers
> Gabriel