Decimating micrographs
Forums
Hello Niko,
Using a fairly high mag (59 000 on film) my Nyquist frequency is 2.4 Å. To calculate the defocus value, the info between 2.4 Å and ~5 Å is more detrimental than useful. I see that you can limit the estimation to a certain resolution range. But is it possible to actually reduce the image size by a factor of two so that I get a better signal and better sampling in the useful range and a speedup of a factor of 4 in computing time?
Is it a good thing to do? or would bad things happen if I do it that way? Maybe it's an option that I didn't see..
By advance, many thanks.
There is an option in CTFTILT
There is an option in CTFTILT to decimate the input image internally. The decimation factor is set by the last parameter on the first input line.
If you do not need to use the high-resolution data in your fit you can decimate your image. This will cause some aliasing of the higher frequencies but if they are not too strong this should not be a problem. We usually run CTFTILT with a decimation factor of 2. CTFFIND3 does not have this option and you will have to decimate your image before running the program. Make sure that you adjust the pixel size accordingly (for example, double it if you do a 2 x 2 decimation). However, CTFFIND3 normally runs so fast that decimation is not necessary. Decimation should not increase the signal in your data but in some cases the Thon rings become less visible. This might be due to the way the background is subtracted in CTFFIND3 and CTFTILT. The background is determined as a low-pass filtered version of the spectrum. If the Thon rings become very broad the filter will not remove the Thon rings efficiently and they will remain part of the subtracted background. A more sophisticated procedure could do better, of course.