[Chimera-users] On the fly map sharpening?
goddard at sonic.net
Tue Jul 21 12:21:26 PDT 2015
Thanks for the suggestion.
I just added an “invert” option to the Chimera 1 "vop gaussian” command that allows trying some map sharpening, for instance, “vop gauss #0 sdev 0.5 invert true”. The normal “vop gaussian” behavior is convolution with a Gaussian which smooths a map. That is equivalent to attenuating the Fourier components by multiplying by a Gaussian. WIth the invert operation it instead amplifies the Fourier coefficients by one over the Gaussian scaling. This only works well for Gaussian standard deviations that are small compared to the map grid spacing (ie pixel size), with too large values producing a map of all speckles. The problem is that the Gaussian attenuation will scale some of the Fourier components to near zero if you use larger standard deviation values. When you try to scale those back up the limited floating point precision in the calculations produces enormous Fourier components. Using standard deviation equal to half of the pixel size worked fine (max attenuation 0.574) but with standard deviation equal to twice pixel size the maximum attenuation is 10e-8 and invert just produces a mess of speckles. Probably the scaling should not be so extreme with an ideal Gaussian and exact arithmetic but the Chimera code is truncating the Gaussian beyond 5 standard deviations and computing Fourier transforms on larger grid sizes to avoid periodic wrapping of the convolution. In any case, this isn’t an ideal way to sharpen a map, but it is a start, in tonight’s daily build.
It will be hard to get “instant feedback” trying sharpening because on a large map these computations can take many seconds.
> On Jul 21, 2015, at 7:26 AM, Oliver Clarke <olibclarke at gmail.com> wrote:
> Hi, just a request for a future chimera version (probably Chimera 2 I guess) - would it be possible to integrate on the fly map sharpening into chimera, perhaps as an option for vop? It would be great to be able to play with different negative B-factors within chimera and thereby get instant feedback on map contrast and detail in different regions.
> Alternatively, would it be possible to call an external program that does the same thing from within chimera, perhaps bfactor from the Grigorieff lab or similar?
> Chimera-users mailing list
> Chimera-users at cgl.ucsf.edu
More information about the Chimera-users