[Chimera-users] gaussian filter
Tom Goddard
goddard at sonic.net
Mon Apr 30 10:49:28 PDT 2012
Hi Mike,
g.step is a 3-tuple giving the grid plane spacing, for example g.step
= (3.5, 3.5, 3.5) means the grid planes are separate by 3.5 Angstroms
along x, y, and z axes. The line of code you are looking at converts
the Gaussian linewidth in Angstroms to grid index units. The g variable
holds a Grid_Data object defined in
share/VolumeData/griddata.py
The code that actually calculates the Gaussian filtering is
share/VolumeFilter/gaussian.py
It does a convolution with a Gaussian using a Fourier transform to speed
it up. It can do cyclic or non-cyclic convolution -- meaning the map is
considered to be periodic or not periodic. This effects filter results
near the edge of the map. In the cyclic case the Gaussian centered near
an edge wraps around to the opposite map box edge. For the non-cyclic
case it behaves as if the map is zero outside its box. The code uses
the Python numpy library to do the Fourier transforms and uses efficient
FFT sizes since this can make an large difference (10x) in the speed
with powers of 2 generally being best. Let me know if you want more
explanation -- I wrote that code.
Tom
> Hi,
>
> I've been using the gaussian filter for a number of years to help me display my noisy data. I'd like to understand what it's doing a bit better, and so I need help with the following code:
>
> ijk_linewidths = map(lambda step: linewidth / step, g.step)
>
> what is the resulting length of ijk_linewidths? I can't seem to figure out what type of data type g.step is. Is it a feature of the image itself, or where does it come from?
>
> Best Regards,
>
> mike
>
>
>
> _______________________________________________
> Chimera-users mailing list
> Chimera-users at cgl.ucsf.edu
> http://plato.cgl.ucsf.edu/mailman/listinfo/chimera-users
>
More information about the Chimera-users
mailing list