Background
The most commonly used assessment of resolution is Fourier ring correlation (FRC) in two dimensions, or Fourier shell correlation (FSC) in three dimensions:
where the sums are over all of the voxels in the shell at spatial frequency s. The resolution is reported at the furthest point where the curve drops below a threshold, typically 0.5 or 0.3.
A second measure is the differential phase residual (DPR):
where f(s) is the phase of a structure factor.
Estimating overall/global resolution
Two reconstructions have to be generated from mutually exclusive sets of particle images:
breconstruct -v 3 -resol 15 -mult 2 -rescale 0,1 -sym D5 -recon klh_run1.pif klh_run1_sel.star
These reconstructions are then compared by FSC and DPR:
bresolve -v 7 -resol 15 -Post klh_run1_resol.ps -map klh_run1_01.pif klh_run1_02.pif.
The postscript output file contains two curves, for FSC and DPR:
Estimating local resolution
The local resolution estimate is calculated for each voxel, using a region around the voxel and subsampling it in reciprocal space:
blocres -v 1 -size 10 -taper hann 36016_2_01_e.pif 36016_2_02_e.pif 36016_2_locres.pif
The -size option gives the size of the box used to calculate a resolution estimate for a voxel. The box should be big enough to avoid high frequency components that would give an artificially high resolution, but small enough to give a representative local measure. In general the median local resolution should be similar to the global resolution.
This program is very computationally intensive and if used as indicated above would calculate many voxels that are typically outside the region of interest. To indicate a smaller region over which to do the calculation, a mask can be generated with bmask and blocres used with the -Mask option to limit the region over which the calculation is done:
bmask -v 7 -rad 30,80 -create 200,200,200 -ori 100,100,100 shell_mask.pif
blocres -v 1 -size 20 -Mask shell_mask.pif -taper hann 36016_2_01_e.pif 36016_2_02_e.pif 36016_2_locres4.pif
In addition, the -symmetry option limits the calculation to an asymmetric unit. The asymmetric unit is then replicated to complete the output image.