[Insight-users] "itkVectorFuzzyConnectednessImageFilter.txx".

[Ying Zhuge]

Hi Gabriele,

That paper is for scalar image case only.
In order to handle vector image, we always do scale-based filtering 
first, that is distance weighting average within the scale, and then 
estimate the homogeneity-based affinity and object-feature-based 
affinity. If we do the way that paper described, it will take much more 
computation time.

m_CirclePointsLoc is a lookup table, which keep the offset location for 
every point within a certain scale region Bk(c) - Bk-1(c). Please note 
there is a print mistake in the equation 4.1 in that paper you 
mentioned,where d should belong to Bk(c) - Bk-1(c).

anisotropy_row and anisotropy_col mean pixel width and height.

count_obj and count_nonobj are used for estimating the fraction FOk(c), 
which equals to count_obj/(count_obj+count_nonobj).

The value of mean_neighbor depends the dimension of your imput image, 
for 2d case, it is the mean value of its 4 neighbors.


Hope these help, and let me know if you have further questions.

Ying

Gabriele Martino wrote:

> Hi,
>
>  
>
> id like to know how itk implements scale-based fuzzy connectedness.
>
> Im sorry if my questions are so stupid but im new in image processing 
> and itk.
>
>  
>
> Ive read this paper, as suggested :
>
>  
>
> " P. K. Saha, J. K. Udupa, and D. Odhner,
>
>   Scale-based fuzzy connected
>
>  image segmentation:
>
>   theory, algorithms, and validation, Computer Vision
>
>     
>
>   and Image Understanding 77, 2000, 145-174. "
>
>  
>
> In this paper is presented a particular form of
>
> both homogeneity-feature-based affinity
>
> and object-feature-based affinity.
>
>  
>
> As for the object component, a filtered value of
>
> a pixel is computed.Ive seen,if im not wrong,
>
> that this component is computed by the method
>
> "computerfilter()" in "itkVectorFuzzyConnectednessImageFilter.txx".
>
>  
>
> As for the homogeneity component, a value,for each couple of pixels,
>
> is calculated based on two digital balls centered on c e d,that is
>
> the couple of pixels.
>
> But i dont unterstand where scale-based-homogeneity, as presented
>
> in the paper below,is computed in 
> "itkVectorFuzzyConnectednessImageFilter.txx".
>
>  
>
> Finally i dont unterstand how scale is computed.
>
> Does m_CirclePointsLoc[x][y][z] contain all the pixels that belong to 
> the digital
>
> ball centered in x,y ?
>
> What does anisotroy_row represent?and anisotropy_col?Why MAXSCALE+5?
>
>  
>
> What is count_obj and count_nonobj?
>
>  
>
> is mean_neighbor a mean of 4-neigh pixels?
>
>  
>
> Cuold you give me a brief description of all that?
>
>  
>
> i hope you'll answer all my questions.
>
>  
>
> Thanks a lot.
>
>------------------------------------------------------------------------
>
>_______________________________________________
>Insight-users mailing list
>Insight-users at itk.org
>http://www.itk.org/mailman/listinfo/insight-users
>  
>