Regional image similarity criteria based on the Kozachenko-Leonenko entropy estimator

Mutual information is one of the most widespread similarity criteria for multi-modal image registration but is limited to low dimensional feature spaces when calculated using histogram and kernel based entropy estimators. In the present article we propose the use of the Kozachenko-Leonenko entropy estimator (KLE) to calculate higher order regional mutual information using local features. The use of local information overcomes the two most prominent problems of nearest neighbor based entropy estimation in image registration: the presence of strong interpolation artifacts and noise. The performance of the proposed criterion is compared to standard MI on data with a known ground truth using a protocol for the evaluation of image registration similarity measures. Finally, we show how the use of the KLE with local features improves the robustness and accuracy of the registration of color colposcopy images.