Registration Using Segment Intensity Remapping and Mutual Information

A method for processing images prior to normalized mutual information based registration and an enhancement of the registration measure are presented. The method relies on k-means clustering of the intensity distribution and a gray-value remapping of spatially unconnected segments of similar intensity. Registration experiments using binary images from the MPEG-7 dataset and retinal images show our preprocessing to cause a significant reduction of the number of misregistrations without loss of accuracy or an increase in optimization steps.

• Type: Conference
• Year: 2004
• Where: MICCAI
• Authors: Zeger F. Knops, J. B. Antoine Maintz, Max A. Viergever, Josien P. W. Pluim
• Keywords: Intensity Distribution, Medical Imaging, MICCAI 2004, Normalized Mutual Information, Registration Measure