Biomedical Engineering Reference
In-Depth Information
multiresolution approach for CT-MR registration using normalized mutual in-
formation. Before the registration starts, the background is segmented in both
images using region-growing. Then we performed one of three binning tech-
niques on the foreground details, while the background is put into one bin.
Three binning techniques were investigated in this manner.
Our results have shown that all three approaches can reach a subvoxel accu-
racy with no loss of speed. The approach using the nonlinear binning technique
shows improvement in accuracy and speed, compared to the other two binning
techniques, since it can achieve less dispersion in the joint-histogram computa-
tion.
