Classification Matrix Image Files

These image files are normally produced by programs like e2classify.py or e2classifykmeans.py. They contain information associating each of a set of particles with one or more reference particles, and by extension, 3-D orientations. They are generally produced (in some cases trivially) from a similarity matrix, by searching it for minimum (best matching) values. This file format permits each particle to (optionally) be associated with more than one reference image. The file normally contains a set of 6 images.

The first image in the file contains the classification information. The size of the image in y is the number of particles that have been classified. The size of the x-axis determines the maximum number of different references each particle can be associated with. Most often nx = 1, meaning each particle is associated with a single reference. While the pixel values (like all EMAN2 images) are floating-point, they actually contain integers indicating reference image number.

The file will normally contain 6 images:

• Image 0 - which 'class' each particle (y axis) is associated with. Multiple values along x- axis
• Image 1 - a 'weight' associated with each classification. This could be used for example, for something like maximum liklihood, so a particle has different probabilities of being in different classes, normally 1
• Image 2 - dx, translation to bring particle in alignment with this reference
• Image 3 - dy, same
• Image 4 - dalpha, rotation
• Image 5 - 0 or 1, flip required