A Novel Deformable Image Registration Algorithm Based on Improved Active Demons

摘要

The optical flow field-based active demons algorithm is a major technique for nonrigid image registration. Current active demons algorithms for large deformation problems are suffered with low registration precision due to using gradient information of the reference image and moving image only. To avoid the disadvantage of the active demons algorithms, this paper first presents the length and area of connected region as driving forces into the active demons diffusion equation and then proposes a new deformable image registration algorithm by combining the gradient information-based driving forces and connected region-based driving forces. Furthermore, a multi-resolution technique is introduced for robust performance against noise. Experimental results show that the proposed deformable image registration algorithm has better performance than several conventional active demons algorithms.