Affine image registration using a new information metric

We present a new information metric for multimodality image registration. The metric is technically a pseudometric since it satisfies the properties, (i) nonnegativity, (ii) symmetry, (iii) triangle inequality and is (iv) zero if (but not only if) the two image intensities are identical. Information metrics are rarely used in image registration and notably, the widely used mutual information measure is not a metric. Given images A and B, the metric used here is the sum of the conditional entropies H(A/B) and H(B/A). We show that when compared to mutual information which can even become negative in the multiple image case, it is easier to extend our metric to the registration of multiple images. And, after using an upper bound, we show that the sum of the conditional entropies can be efficiently computed even in the multiple image case. We use the metric to simultaneously register multiple 2D slice images obtained from proton density (PD), magnetic resonance (MR) T2 and MR T1 3D volumes and to match human face images obtained wider different illuminations. Our results demonstrate the efficacy of the metric in affine, multiple image registration.