[vtkusers] about vtkImageMask

[David Gobbi]

On Jan 14, 2008 10:20 AM, Sean Ziegeler <seanzig at users.sourceforge.net> wrote:
> >> Could anyone tell me why the mask input in the
> >> vtkImageMask is type "unsigned char"?
> >> I think if it could be bool/bit, it could save a lot
> >> of memory, since we only need values 0 and 1 in a
> >> mask?
> >
> > I'm not a specialist but I would say that booleans are coded on 8 bits
> > (even if they can take only two values). I know no processor capable of
> > handling 1 bit values. So using boolean would _not_ save any memory.
>
> The processor cannot handle single bits directly, but you could write
> code using logical bit operations (bit shifts, bitwise-and's, etc.) to
> pack each bit into a byte and to extract each back out.  It would save
> 8x space, but all the extra ops will probably slow you down by at least
> 8x, if not more.  Also the code would get rather hairy.  All-in-all,
> using bytes is probably better.


If memory is a concern, take a look at the documentation for
vtkImageToImageStencil and the vtkImageStencilFilter.  I added
a special kind of data set to VTK called "vtkImageStencilData"
a while back, the "StencilData" is a run-length compressed binary
image mask. Unfortunately it isn't very well supported yet, but you
can shoot me an email if you have questions about it.

 David