Power spectrum with binned images
Forums
I would like to ask you how ctffind3 generates a power spectrum (or is it an amplitude spectrum) from a large image by using a 512x512 block. Does the program average the power (amplitude) spectra generated as the 512x512 window is scanned over the large image?
The reason why I am asking now is that we found that he obtained a much better looking power spectrum (more Thon rings) if he first binned the original image by 2. Do you already know, by the way, whether initially binning by 2 will actually improve the estimation of defocus and astigmatism, or is it just a "cosmetic" thing, which won't affect CTF estimation?
In another issue, we also tried to run ctffind3 after binning the original image by 4, but then ctffind3 did not work. We think something may have gone wrong when binning by 3, but if you have another explanation, that would be good to know about it.
CTFFIND3 calculated the power
CTFFIND3 calculated the power spectrum (intensities) of tiles of size specified by the user. The tiles are generated by cutting up the original image. There is no overlap between titles. The power spectra of all the tiles are then averaged.
There is usually a lot of background in these power spectra that is not modulated by the CTF and needs to be removed before fitting the CTF. This is done in CTFFIND3 by low-pass filtering the original power spectrum using a box convolution (the boxes are adjusted at the edges of the power spectrum “image” so that edge effects are minimized). This low-pass filtered version is then subtracted.
CTFFIND3 adjusts the box size of the box convolution according to the original pixel size/magnification of the input image so that the filter does not include the CTF modulation (which we want to fit in a later step). The adjustment of the filter is pretty crude and does not always work. It most likely is the explanation of why the power spectrum looks different with different binning of the input image.
The filtering (and other things) can be improved in CTFFIND3. I have not had time to do so, unfortunately.