Image matching by correlation analysis

Date of Award




Degree Name

Doctor of Philosophy (Ph.D.)

First Committee Member

Joachim H. Nagel, Committee Chair


The diagnostic potential of medical images obtained at different times or from different imaging modalities can be augmented by objective, accurate matching of the different data sets. This dissertation describes the development and testing of an algorithm in which the matching is done in the Fourier domain. No previous edge or feature detection is required.Initially the images differ in translation, rotation, and scale. These parameters are coupled, thus they cannot be computed simultaneously. The algorithm creates a sequence of intermediate representations of the images. Each representation is invariant to an additional matching parameter. The matching variables are computed, one at a time, from these intermediate, invariant image representations. First the scale factors are computed from rotation and translation invariant representations of the input images. Next the rotation angles are computed from translation invariant representations of scale equivalent images. Then the translation distances are computed from the scale and rotation equivalent image data.The theory is described in three dimensions. Testing was done using a two dimensional implementation. Results are presented from experiments using synthetic, phantom, and patient image data. The resulting computationally inexpensive algorithm is an objective, accurate, and general approach to the image matching problem.


Engineering, Biomedical

Link to Full Text