Filter_Projections

filter_projections

Purpose

filter_projections allows you to discard images that do not complain with a quality criteria. This program is thought to be integrated in an iterative reconstruction protocol in which all the images are treated separately and some of them may be discarded due to their low quality. The algorithm may also put a mask and renormalize according to a reference volume.

Usage

$ filter_projections ...

Parameters

• `` The documentfile filename with the list of projections as it is output by the alignment programs
• `` This rootname will be used to generate a .doc and a .sel with the list of selected images
• `` The Score is the quality measure ofAngular_predict. This filter will remove images which are in the lower specified percentil.
• `` The Cost is the quality measure ofAngular_predict_continuous. This filter will remove images which are in the higher specified percentil.
• `` This filter will remove images that have moved more than certain limits with respect to the assignment in the previous iteration (docfile0).
• `` This filter removes images whose correlation with the reprojected images from the volume are in the low percentil. The correlation is only measured within r1. This filter also produces a set of images (rewriting the input ones) whose grey values have been adapted to the reference volume. r2 is used to put a cosine raised mask between r1 and r2.
• -thr [N1] = The filter on normalization may be slow since many reprojections must be performed. If several processors are accessible for threading, give with this option the number of processors.

Examples and notes

$ filter_projections -i Iteration01/angles01b.txt -o preproc_recons -thr 2 
     -filter_score Iteration01/angles01.txt 10 \
     -filter_cost Iteration01/angles01b.txt 10 \
     -filter_normalization Iteration01/model00F.vol 32.0 36.0 10

--Main.CoSS - 18 Dec 2008