IJPAM: Volume 37, No. 1 (2007)


Shaoyan Sun$^1$, Liwei Zhang$^2$, Chonghui Guo$^3$
$^{1,2}$Department of Applied Mathematics
Dalian University of Technology
Dalian, 116024, P.R. CHINA
$^1$e-mail: sunshaoyan1979@yahoo.com.cn
$^2$e-mail: lwzhang@dlut.edu.cn
$^3$Institute of System Engineering
Dalian University of Technology
Dalian, 116024, P.R. CHINA
e-mail: guochonghui@tsinghua.org.cn

Abstract.A new histogram estimation scheme is proposed in the calculation of mutual information (MI) for image registration. We show that the mutual information function derived from the new estimation scheme is continuous and differentiable, and we derive analytic expressions of its derivatives that allow numerically exact evaluation of its gradient. Both Powell and conjugate gradient methods are applied to optimize the MI function. Experimental results show that, compared with the MI using partial volume interpolation (PVI), the MI registration function derived from the suggested scheme reduces the interpolation-induced artifacts, encounters less local maxima and the registration performance is improved significantly.

Received: January 23, 2007

AMS Subject Classification: 94A15, 94A08, 90C90

Key Words and Phrases: image registration, mutual information, joint histogram estimation, Powell's method, conjugate gradient method

Source: International Journal of Pure and Applied Mathematics
ISSN: 1311-8080
Year: 2007
Volume: 37
Issue: 1