mag_distortion_correct results


I was wondering how to verify the mag_distortion_correction result.

I did mag distortion estimation on an original image and the results were:
distortion angle 42.0
major scale 1.015
minor scale 0.985
total distortion 2.97%
stretch only parameters: 42 / 1.03 / 1 / 0.626

After doing mag distortion correction with the stretch only parameters, I did mag distortion estimation again on the corrected image, and the results were:
distortion angle 42.2
major scale 1.007
minor scale 0.993
total distortion 1.40%
stretch only parameters: 42 / 1.014 / 1 / 0.622

Also the pixel size (from the header info) of the corrected image shows 0.1348E-03 for x and 0.1303E-03 for y, which I assume should be 0.626.

Any ideas?



Hi Huaibin,

One thing I should check, is that your input is a stack of images of 10-20 different areas of gold. One image of gold is not sufficient for the estimation. When you rerun on the corrected image, do you mean you are correcting the 10-20 input images and rerunning? How many different areas do you have in your input stack?

If you are running on a good size stack of different areas - if you look at the amplitude spectra you can get an idea for how the correction has gone. The program outputs the calculated amplitude specta from the original data (this defaults to my_amp.mrc), then the rotational average of that (my_amp_rot.mrc). If you overlay these there should be an obvious distortion when flicking between them (the my_amp_rot.mrc is by definition rotationally symmetric). my_amp_corrected.mrc is the original amplitude spectra corrected and if you overlay this with the rotational average there should be no obvious distortion.


In reply to by timgrant

Hi Tim,

I was using an image stack of 10 images of different cross grating grid area.
I think I figured out what the problem is.
I was using non-expert mode for distortion estimation, and the default Max scale factor is 0.03. I then used expert mode and changed the Max scale factor to 0.05, and the total distortion found was 4.55%. After distortion correction with these new parameters, I ran estimation on corrected image stack, and the new distortion found is ~0.1%.

I have two more questions
1. Is it normal that the distortion at 29kx would be as big as 4.55% for an F20? I also did the estimation for 44kx and the total distortion is 3.58%.

2. We have a lot of data taken under 29kx with K2 counting mode (image pixel size is 1.27A). When doing distortion correction with mag_distortion_correct, how severe the interpolation would affect the final structure resolution? We are not aiming at resolution of 4A or better for now.


In reply to by hwang

Hi Huaibin,

Glad you've gotten to the bottom of it!

4.55% is the highest distortion I have heard of, seems like your microscope is particularly bad in this respect.

The interpolation shouldn't be that bad really, it will be decreasing your signal a reasonable amount at 2.6A (your Nyquist), but shouldn't really affect 4A. If you are worried, you can pad your images x2 in Fourier space (i.e. resample to 0.635A per pixel) prior to the correction, and then cut them back down again after the correction, if you do this, there should be no real affect even at 2.6A.