Structural Study of Heterogeneous Biological Samples by Cryoelectron Microscopy and Image Processing

Figure 3

Eigenimages and Classification. (a) A set of raw images. (b) Four images (top) shown with a coarse pixilation similar to those in panel (a) with size pixels. Images form a matrix where a single image is presented as a single row in it (bottom). Each pixel in row 1 of image 1 is laid out in the first row of the matrix. The second row of image 1 follows on after row 1 in the first row of the matrix. This continues until all rows have been laid out in the first row of the matrix. The rows of image 2 are laid out in a similar manner in row 2 of the matrix and the process continues until all images in the dataset have been placed into the matrix. (c) Eight eigenimages obtained from the set of aligned images in (a). (d) Classification of the dataset into 5 classes. (e) Classification of the dataset into 10 classes. (f) Raw unaligned rotated images. (g) Eigenimages from the unaligned dataset.

(a)

(b)

(c)

(d)

(e)

(f)

(g)