Per Particle Micrograph corrections

[garrettwrong]

Amit passed along the following from a Flatiron meeting relating to an alternative to picked particle masking. The main idea is that we know CTF corruption spreads out beyond the actual particle location, and where there are dense concentrations of particles these neighbors' corruptions can combine with the central picked particle.

Consider a micrograph that has already been picked and used for an initial reconstruction. The reconstructed volume along with estimated poses could be used to generate a stack of simulated particles mapping to the original picked particles. The estimated CTFs can be applied to these simulated particles. Then, for each of the original picked particles, for each local neighbor, the relevant CTF corrupted contributions can be subtracted from the original micrograph content, hopefully yielding a more accurate representation of the central picked particle.

There are at least two ways to approach this by leveraging existing tools within ASPIRE-Python. One is to consider the application on a per particle basis (ie, as outlined above).

Another is to create whole simulated micrographs of estimated particles, then apply the CTFs to create simulated ctf corrupted micrographs. Subtracting the sim ctf corrupted micrograph from the original micrograph yields the aggregation of removing all simulated picked particles (and their ctf corruptions). At this point, for any particle, we can add back that sim ctf corrupted particle and the original particle to generate the original particle with all other simulated contributions on that micrograph removed. I suspect this later approach would be more computationally efficient.

It was suggested to keep this around for potential intern/student project in the future.