Date of Graduation

2004

Document Type

Thesis

Degree Type

MS

Committee Chair

Hany Ammar

Abstract

In the development of Automated Dental Identification System (ADIS), we focus on registration to align the subject dental regions of interest with respective reference regions. With the goal to achieve high performance for dental image registration, a dualstage approach is proposed which is based on the hypotheses generation-verification paradigm. The registration of the dental regions of interest is the basis for hypotheses generation. The hypotheses generated by using a fast method are verified to retrieve the candidate parameter sets to give accurate alignment of the regions. The statistical distributions of local geometric properties of image contours constitute the essential concept for achieving registration of reference and subject images. Affine transformation is applied to the subject image to get it best aligned to the reference image. The statistical method accomplishes the task of generating hypotheses for registration without the need of feature correspondences thus reducing the computational time. The generated hypotheses are verified by considering mutual information between all the registered image pairs. Establishing pixel-to-pixel correspondences of the images is time consuming and leads to computational problems. But speed is not affected as we execute our verification process only for a short list of potential candidates generated from the first stage. With mutual information being an accurate and reliable measure, the performance of the image registration system improves. With the efficient integration of the stages, our proposed approach for dental image registration gives the dual benefits of speed and accuracy.

Download

Share

COinS