Use this URL to cite or link to this record in EThOS:
Title: Registration of optical images to 3D medical images
Author: Clarkson, Matthew John
Awarding Body: University of London
Current Institution: King's College London (University of London)
Date of Award: 2000
Availability of Full Text:
Access from EThOS:
Access from Institution:
The work described in this thesis deals with the registration of single and multiple 2-dimensional (2D) optical images to a single 3-dimensional (3D) medical image such as a magnetic resonance or computed tomography scan. The approach is to develop an intensity based method using an information theoretic framework, as opposed to the more typical feature or surface based methods. Relevant camera calibration and pose estimation literature is reviewed, along with medical 2D-3D image registration. An initial algorithm is developed, which performs registration by iteratively maximising the mutual information of a rendered image and a single optical image. The framework is extended to incorporate information from multiple optical and rendered images which signi cantly improves registration performance. A tracking algorithm is proposed, which augments this framework with texture mapping as a means of achieving alignment over a sequence of optical images. These methods are tested using images of skull phantoms and volunteers. A new measure based on the concept of photo-consistency, used in the surface reconstruction literature, is proposed as a measure of image alignment. The relevant theory is developed. This new method is tested using a variety of different photo-consistency based similarity measures, optical images, different numbers of images, images with varying amounts of added noise, different resolutions and different camera positions relative to the object of interest. In almost all cases, similarity measures based on this new framework perform accurately, precisely and robustly. Potential applications will be in radiotherapy patient positioning, image guided craniofacial, skull base and neurosurgery, computer vision and robotics, where the accurate alignment between a 3D image or model and multiple 2D optical images is required.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available