Combined Local and Global Image Registration and its Application to Large-Scale Images in Digital Pathology

A large-scale, nonlinear image registration problem can be partitioned into smaller independent subproblems by adding a global, coarsely discretized distance measure. The remaining inconsistencies between subdomains are smaller than without the coarse distance term and can be incorporated into the global solution by a blending method. Reaching a similar accuracy, the new method enables the registration of large-scale images that could otherwise not be computed.