An Improved Medical Image Registration Framework Based on Mutual Information

This paper presents an improved framework for medical images registration. Comparing with the previous registration framework, this framework uses the Mutual Information (MI) as main measure method and is more precise in image registration. Aside the input and output data, the framework can be separated into four parts: interpolator, measurer, optimizer and transformer. Interpolator is used for evaluating moving image intensities at non-grid positions. Measurer provides an appraisal method of how well the fixed image is matched by the transformed moving image. Optimizer can optimize the measure criterion and transformer exerts some transformations on the objective image. Measurer component is the most critical element of the framework and we adopt Mutual Information as our main measure method. These four parts act as different roles in medical images registration and construct a simple, accurate and stable medical images registration framework. We have already realized the framework and got a satisfying result.