compile for big images

Forums

Hello,
I am wondering if I have to compile frealign in order to process images bigger than 256. The current documentation suggests that I do, but I would prefer to avoid compiling if I don't have to.

Thanks,
Scott

Hi Scott,

If I'm not mistaken the frealign version currently available on the website (here; v. 8.08) uses dynamic memory allocation, which should mean it can deal with arbitrary image sizes, without the need to recompile.

HTH
Alexis

Hi Scott,

I don't know if the question is still up-to-date, but we have just installed Frealign in our lab and tested it with images of 368 pixels of edge length. Everything is working perfect! The only problem we had in the beginning was the padding factor, which was set to 4 and blew away the memory of our small test system. Setting it to 1 got rid of the problem, though. Other than that everything is working great and really fast.

Best regards,

Michael

In reply to by adesgeorges

Hi Amed,

here's how I understood the advantage: If you have an image of 256 pixels edge length, the spacial frequencies will be distributed along 128 pixels in fourier space. If you now pad your image to 512 pixels (in the case of pad factor = 4 it would be 1024), you have got 256 pixels to distribute your frequencies on, therefore a finer separation of frequencies and, hence, a more precise alignment.

However, I would very much appreciate it, if other forum users could back me up on this.

Best regards,

Michael

In reply to by msaur

Yes, this is correct. The projections calculated internally will contain fewer interpolation errors. Also, the interpolation is faster because only the nearest neighbors are considered when calculating a projection. Therefore, the refinement runs faster. However, if you do not have sufficient memory you should not worry. I have never seen a resolution noticeably limited by interpolation errors in the matching projections. The main difference will be the slower speed.