Issue
When comparing raw cross-correlation values for the same particles between the Leopard-EM match_template and refine_template program, there are meaningful differences between the two. The main result of this issue is that program results become hard to quantitatively compare although results within each program are self-consistent.
These differences arise from the refine_template program using local information from the extracted particle boxes to calculate the whitening filter while the match_template program uses the global power spectra to filter the image.
Potential fixes
- Only use global whitening within the
ParticleStack class when calculating the Fourier filters
- Add argument to switch between local/global whitening within the particle stack
Both of these options should be relatively easy to implement within the ParticleStack class, and would expect the raw cross-correlation values remain consistent (should test this).
Issue
When comparing raw cross-correlation values for the same particles between the Leopard-EM
match_templateandrefine_templateprogram, there are meaningful differences between the two. The main result of this issue is that program results become hard to quantitatively compare although results within each program are self-consistent.These differences arise from the
refine_templateprogram using local information from the extracted particle boxes to calculate the whitening filter while thematch_templateprogram uses the global power spectra to filter the image.Potential fixes
ParticleStackclass when calculating the Fourier filtersBoth of these options should be relatively easy to implement within the
ParticleStackclass, and would expect the raw cross-correlation values remain consistent (should test this).