MI image registration using prior knowledge

Subtraction of aligned images is a means to asses changes in a wide variety of clinical applications. In this paper the information theoretical origin of Mutual Information is explored. The parallelism between the problem of image registration and its characterization as a communication channel is found to be limiting. The concept of Mutual Information (MI) is extended to Focussed Mutual Information (FMI) to incorporate prior knowledge. Several methodologies are proposed to address specific registration problems, incorporating FMI. These methods are demonstrated by means of some elucidating examples.