[SciPy-user] ndimage starting points

David Warde-Farley dwf@cs.toronto....
Thu Oct 23 03:08:35 CDT 2008

Hi Stefan,

Thanks for the quick reply!

On 23-Oct-08, at 3:39 AM, Stéfan van der Walt wrote:

> Hi David
> 2008/10/23 David Warde-Farley <dwf@cs.toronto.edu>:
> The ndimage module comes from Numeric.  Unfortunately, the version we
> included came from a different source and did not have all the Numeric
> documentation.  It would be great if you could write a patch to bring
> some of the docs over -- they may also answer your questions below.

Did you mean to say numarray? Jarrod pointed me at the docs for a  
numarray nd_image module. Hopefully there isn't also a module from  
Numeric as well...

Come to think of it, I do know why I missed those docs -- Google isn't  
smart enough to return results for "nd_image" when you search for  
ndimage :)

>> - Some things like variance() don't immediately seem to add anything
>> to the standard numpy functions, I am assuming that the ability to
>> mask by label is their key advantage. Can someone confirm or correct
>> this?
> Much of the ndimage functionality was implemented before NumPy.  I
> think we can already replace a big part of its functionality using
> Python + Numpy, without going down to the C level (and I think this
> would benefit the library in general).

So, the one thing that, for example variance() does is take a 'labels'  
parameter and an 'index' parameter, which seems to suggest to me that  
the purpose of this is to compute variance within a labeled object (or  
all labeled objects, disregarding the background). I imagine this  
would be easily reimplemented in pure Python, though, you're right.

For now I'll see about getting this documentation up to snuff using  
the magical tome Jarrod pointed me toward.



More information about the SciPy-user mailing list