Image Registration is an important tool for medical image analysis. It can be used to find complex spatial relationship between images even in the case of multi-modality data. Point similarity measures were shown to provide several benefits to the registration process. Although they are intensity-based, they enable multi-modality assessment of the most localized image discrepancies by measuring similarity of arbitrary small image subregions, including individual image points. Such local properties enable registration process to avoid interpolation artifacts commonly observed using other intensity-based similarity measures. Furthermore, point similarity measures separate the registration process into functionally independent parts of similarity measurement, optimization and spatial regularization, simplifying design and testing of registration methods. Finally, they enable straightforward integration of additional knowledge of the problem domain, and thus enable additional registration improvements.