We have investigated an intensity-based image registration technique using a robust correlation coefficient as a similarity measure. The proposed method has an advantage over the ordinary correlation coefficient since it reduces the effect of "outlier" image intensity values. For the application of image registration to radiotherapy or image-guided surgery, there may be outlier samples due to the presence of the objects such as surgical instruments. We have verified the usefulness of the proposed method by simulation and phantom experiment.
Jeongtae Kim, Jeffrey A. Fessler