Particle-guided Image Registration

摘要: We present a novel image registration method based on B-spline free-form deformation that simultaneously optimizes particle correspondence and similarity metrics. Different from previous methods optimized w.r.t. the control points, in our is estimated set of dense unstructured pair which we refer as corresponding particles. As intensity values are matched location, performance iteratively improved. Moreover, use particles naturally extends to group-wise by computing mean Motivated surface-based method, developed system takes such domain, while keeping spirit similar. The core algorithm both minimizes an entropy metric well maximizes space sampling demonstrate results application rodent brain structure segmentation show provides better accuracy two structures compared other methods.