[Insight-users] Suggestions for the best Data Structure

[Luis Ibanez]

Hi Arash,

It seems redundant to have the coordinates
for all the pixels represented explicitly.

If you have a 2D to 3D transformation allowing
to compute (x,y,z) from (u,v) values, you could
avoid taking all the memory for storing such
information on a pixel by pixel basis.

You probably can represent this as a type

     itk::Image< itkRGBPixel<char>, 2 >

and an ITK transform.

----

Unless your image is composed of scattered
pixels, or its mapping in space is given
by a deformation fields, it is probably not
worth to represent this data in the form
of a Mesh.


----

Regards,


    Luis


------------------------
Arash Jahangir wrote:
> Hi all,
> I have been reading this thread for a few weeks and have been impressed 
> with the quality of the people on this thread.  I am hoping that your 
> knowledge can help me decide what basic data structure I should use for 
> manipulating my input data.  I have an input file format that can best 
> be described as bitmap image plus the real world coordinates for each 
> pixel.  The format is:
>  
> u, v, r, g, b, x, y, z
>  
> where (u, v) pair define the row, column coordinates of the pixel and 
> the x,y,z are the real world coordinates.
>  
> I will need to perform all the usual activities such as image to image 
> and image to object registration as well as other image and object 
> modifications/edits.
>  
> The question in my mind is whether I should load the input into two 
> separate data structures of appropriate image and mesh types, Or should 
> I create an ImageSpatialObject (BloblSpatialObject?)  that holds all the 
> data and then extract the mesh and/or image components from it as needed.
>  
> your help is greatly appreciated,
> Arash
>