Global Optimization of Medical Image Registration Based on Nonlinear Correlation Measurement

A global optimization technique for image registration using the concept of nonlinear correlation information entropy (NCIE) as the matching criterion is presented. The method makes it possible to efficiently overcome the local minima problem utilizing the extremum property of NCIE. Furthermore, the improved downhill simplex algorithm incorporated variant accuracy tolerance can reduce the evaluation times greatly as well as obtain an accurate and optimal transformation. Applying to register the magnetic resonance images of brain, the global optimization technique shows the validity and robustness in global searching range.