A Survey on Non-Rigid Image Registration

PhD Qualifying Examination


Title: "A Survey on Non-Rigid Image Registration"

Mr. Shu LIAO


Abstract:

Image registration is a task to reliably estimate the geometric 
transformation such that two images can be precisely aligned. It plays an 
important role in medical image analysis, group analysis and statistical 
parametric mapping. Particularly, medical image registration has attracted 
many researchers over the past decades and become an active research 
topic. Medical image registration can be broadly classified into two 
categories: rigid and non-rigid image registration. They are classified 
according to the transformation model adopted. More precisely, rigid image 
registration uses the rigid transformation as the transformation model and 
therefore it has six degrees of freedom. Non-rigid image registration 
transforms one image to the other image via deformable transformation 
models such as B-spline basis function and free form deformation. 
Non-rigid image registration is more challenging than rigid image 
registration as it has much higher degrees of freedom.

In this survey, we review and analyze three major classes of previous 
approaches proposed for non-rigid medical image registration: landmark 
based, intensity based and feature based non-rigid image registration 
methods. Lankmark based non-rigid image registration methods use prior 
knowledge obtained from manually located landmark points on the input 
images, features are extracted from these landmark points to determine the 
optimum transformation. Intensity based non-rigid image registration 
methods define a similarity measure function with respect to the intensity 
distributions of input images, the value of the similarity measure 
function guides the registration process. Feature based non-rigid image 
registration methods use feature vectors as signatures for each voxel, the 
registration problem thus is formulated as a feature matching and 
optimization problem. We discuss the advantages and disadvantages of each 
class of methods and finally investigate some possible directions in 
non-rigid medical image registration.


Date:     		Wednesday, 25 February 2009

Time:                   3:30p.m.-5:30p.m.

Venue:                  Room 4483
 			lifts 25-26

Committee Members:      Dr. Albert Chung (Supervisor)
 			Dr. Chiew-Lan Tai (Chairperson)
 			Dr. Huamin Qu
 			Prof. Dit-Yan Yeung


**** ALL are Welcome ****