xmipp_image_sort_by_statistics (v3.0)


Sorts the input images for identifying junk particles The program associates to each image a vector composed by the histogram of the image (this accounts for factors such as min, max, avg, and standard deviation, plus a more complete description of the image gray levels) and the radial profile of the image squared.

These vectors are then scored according to a Gaussian distribution that can be chosen to be univariate or multivariate. The multivariate gaussian is more powerful in the sense that it can capture relationships among variables.

If you choose a threshold, you must take into account that it is a zscore. For univariate and mulivariate Gaussian distributions, 99% of the individuals have a Z-score below 3


-i <selfile>
Selfile with input images
-o <rootname=>
Output rootname rootname.xmd contains the list of sorted images with their Z-score rootname_vectors.xmd (if verbose>=2) contains the vectors associated to each image If no rootname is given, these two files are not created
--train <selfile=>
Train on selfile with good particles
--zcut <float=-1>
Cut-off for Z-scores (negative for no cut-off) Images whose Z-score is larger than the cutoff are disabled
Identify also multivariate outliers
Add columns also to input MetaData

User's comments