University of Houston

Gene regulation process has been proved to be not only dependent on the genome sequence but also on the nuclear architecture within a cell. An understanding of this phenomenon requires the analysis of a large number of cells to put them into a common spatial basis. Therefore, algorithms for three dimensional alignment of cells become essential for understanding of nuclear architectures. This thesis evaluates both global and local feature based alignment methods provides a quantitative comparison between obtained results.

The alignment methods used in this thesis are: (1) global feature - PCA based, (2) local features - SIFT based, RACD based and SPIN images based. The alignment methods are compared using a three step technique. The two first steps aim at ensuring the reliability of the different algorithms implemented in providing accurate rotational parameter estimates, whereas the third step consists of application of those methods to align cells from a real dataset. The comparison of the alignment methods indicates that local feature based methods are more accurate than the global feature based method.