Undesired lines after the magnification distortion correction

Submitted by lin on Sun, 2016-02-14 22:45

Forums

Magdistortion

I'm trying to correct the magnification distortion of K2 derived mrc
data, however after executing the /mag_distortion_estimate_openmp_8_18_15.exe file , there are some lines in the processed mrc file , which looks like that the picture have been touched by feline claw.

timgrant Mon, 2016-02-15 09:34

Hi Lin,

Are these lines at the edge of the image, or across the whole image? Also, what parameters did you use for the correction, and how were they estimated?

Thanks!

Tim

lin Mon, 2016-02-15 23:36

These lines are across the whole image, the distortion angle is 136.3, major scale is 1.017, minor scale is 0.998, they are estimated from images of polycrystalline gold. The lines are most obvious when the corrected mrc files are displayed via e2display with the parameter: Mag 0.513 Min 33.1 Max 65.3 Brt 0.296 Cont 0.993 Gam 4.3. However, I can't find these lines from the raw mrc images before correction of magdistortion.

timgrant Tue, 2016-02-16 08:56

If the lines are more like distortions in a grid pattern over the image, then it's probable that you are seeing artifacts due to the linear interpolation used during the correction. I cannot be sure of this without actually seeing an image, maybe you could attach an image demonstrating the lines?

It may be that they are enhanced when displayed in e2display in that way due to addition interpolation that is occurring within the display itself. The artifacts themselves are most powerful close to the Nyquist of the image, and so if your resolution is getting towards Nyquist they may be affecting you, otherwise they are unlikely to do any harm. It is for this reason we correct our distortion on super resolution images, then Fourier bin by a factor of 2.

Finally, did you estimate and correct using the latest versions of the estimate and correct programs? It seems that your major scale and minor scale are not symmetrical which I had thought should be impossible with the latest versions.

Tim

timgrant Wed, 2016-02-17 11:23

Hi,

I guess the grey scaling of those two images is quite different so it is difficult to see from that picture. Could you upload the raw frame before and after correction?

Tim

timgrant Fri, 2016-02-19 21:47

Hi,

I agree that is looks a little strange. I am wondering if it is a combination of the interpolation during the correction, and the interpolation due to the scaling in EMAN.

Would you be able to share the raw frame with me, so that I could investigate further?

Thanks,

Tim

timgrant Mon, 2016-02-22 11:58

Hi,

Thank you for uploading the images - I think I understand what is going on now! The pictures you showed are highly thresholded to show the most positive values correct? When looking at the images at "normal" grey ranges I cannot see any lines, but when I threshold images I do also see them. I'm 99% sure these lines are due to aliases caused by the linear interpolation used for the correction. They have a very small magnitude (which is why you only see them when you use an extreme threshold), and so they probably won't cause you any problems.

If you are worried about them, and would rather not have them, you should be able to mostly remove them by using an upsampled image for the correction. You could do this either by correcting on super-resolution images, or if you don't have super resolution images - first resample your images to 2+ times their original size via Fourier cropping (e.g. using the resample.exe program distributed with Frealign), perform the correction on this image, then downsample the corrected image back the the original size.

Let me know if you still think you are having problems!

Thanks,

Tim