Protocol to perform high-resolution reconstructions
- class xmipp3.protocols.protocol_reconstruct_highres.XmippProtReconstructHighRes(**kwargs)¶
This is a 3D refinement protocol whose main input is a volume and a set of particles. The set of particles has to be at full size (the finer sampling rate available), but the rest of inputs (reference volume and masks) can be at any downsampling factor. The protocol scales the input images and volumes to a reasonable size depending on the resolution of the previous iteration.
The protocol works with any input volume, whichever its resolution, as long as it is a reasonable initial volume for the set of particles. The protocol does not resolve the heterogeneous problem (it assumes an homogeneous population), although it is somewhat tolerant through the use of particle weights in the reconstruction process.
It is recommended to perform several global alignment iterations before entering into the local iterations. The switch from global to local should be performed when a substantial percentage of the particles do not move from one iteration to the next.
The algorithm reports the cross correlation (global alignment) or cost (local) function per defocus group, so that we can see which was the percentile of each particle in its defocus group. You may want to perform iterations one by one, and remove from one iteration to the next, those particles that worse fit the model.
- AUTOMATIC_ALIGNMENT = 2¶
- GLOBAL_ALIGNMENT = 0¶
- LOCAL_ALIGNMENT = 1¶
- NO_ALIGNMENT = 4¶
- SPLIT_FIXED = 1¶
- SPLIT_STOCHASTIC = 0¶
- STOCHASTIC_ALIGNMENT = 3¶
- adaptShifts(fnSource, TsSource, fnDest, TsDest)¶
- calculateAngStep(newXdim, TsCurrent, ResolutionAlignment)¶
- checkInfoField(fnDir, block)¶
- prepareImages(fnDirPrevious, fnDir, TsCurrent, getShiftsFrom='')¶
- prepareMask(maskObject, fnMask, TsMaskOut, XdimOut)¶
- prepareReferences(fnDirPrevious, fnDir, TsCurrent, targetResolution)¶
- readInfoField(fnDir, block, label)¶
- writeInfoField(fnDir, block, label, value)¶
- xmipp3.protocols.protocol_reconstruct_highres.getPreviousQuality(img, imgRow)¶