| 摘要 |
An image registration system and method employs a registration processor (130) to find an optimal registration of two volumetric images (110, 112) by finding a spatial orientation that optimizes the statistical measure cross-entropy. The cross-entropy can be minimized where one or more likely or reasonably good a priori estimations of the true joint probability density function (pdf) of voxel value pairs are available. Where such an estimate is not available, the joint pdf for one or more known incorrect spatial arrangements (e.g., based on prior misregistrations), can be used as the prior pdf estimate(s). Since the prior pdf estimate(s) are unlikely to be the true pdf, a spatial orientation maximizing the cross-entropy is found. Likely and unlikely prior pdf estimations can also be used together to register images. The transformation that optimizes the cross entropy provides an optimal registration, and the parameters for the optimized transform are output to memory (150) for use by a display system (160) in aligning the images for display as a fused or composite image. |
