Accompanying CD-ROM

Videos demonstrating parts of the thesis have been put on a supplementary CD-ROM. This CD-ROM is bound to the back cover of this thesis. It contains illustrative animations that are sometimes referenced in the text. The filenames on the CD-ROM are arbitrary. These are neither chronologically sorted, nor do they correspond to figure/section numbers in the text.

Peripheral files (GIF-formatted):

/Thesis/p1.gif: a video of a combined model built from the MGH data. Seven modes of variation are normally (i.e. drawn from Gaussian distribution) varied simultaneously.
/Thesis/p2.gif: combined model built from the MGH data. Ten modes of variation are normally varied simultaneously.
/Thesis/p3.gif: the registration assessment framework illustrated schematically
/Thesis/p4.gif: chequerboard image of 2 brain images. The sequence shows the chequerboard image as an SSD-based registration proceeds and, in response, the chequerboard image evolves.
/Thesis/p5.gif: one-dimensional registration example. Rows in the matrix represent intensity vectors (1-D images) being registered to align with the remainder of the set in a pairwise fashion.

Larger peripheral files:
(Compressed or uncompressed AVI format; no proprietary codecs required)

/Thesis/p6.avi: sets of 10 1-D vectors are being aligned using multi-edge clamped plate spline (CPS) warps. Ten iterations (systematic passes) through the data are shown, unregistered images on the left and progressively re-registered images on the right.
/Thesis/p7.avi: sets 10 1-D vectors, as visualised in 3-D space, are being aligned. Fifty iterations through the data are shown in the sequence.
/Thesis/p8.avi: the first and second modes of a combined (shape and intensity) model which is automatically built
/Thesis/p9.avi: two 1-D images are being registered. Unregistered images are shown on the left and progressively registered on the right.
/Thesis/p10.avi: 1-D vectors being registered using a model-based objective function
/Thesis/p11.avi: 10 simplified 1-D vectors, which were composed of 4 edges, are being registered. The objective function is based on mean-squared-differences.
/Thesis/p12.avi: 10 1-D vectors are being registered using an objective function that is based on minimisation of the complexity of a point distribution function
/Thesis/p13.avi: a large number of 1-D vectors (visualised as rows) being registered by considering just one vector at a time
/Thesis/p14.avi: automatically-built combined model of a bump. The model is built automatically from the raw training set.
/Thesis/p15.avi: a large-scale illustration of 1-D registration of bump data

Roy Schestowitz 2010-04-05