Today, imaging techniques are essential part of common medical practice. Among these techniques, image registration and fusion plays important role. The task of image registration (or alignment) is to find geometrical transform between two images that eliminates the difference in patient position during image acquisition. The image pair can differ also by the acquisition time, the type of the sensor and imaging method, or images can originate from different patients or atlas data. Image fusion then combines two images into one, so that it is possible to gain maximum information from the two images (an example is the fusion of anatomical MRI image and functional PET image). We are working on new and derived methods of automatic image registration and fusion focused on multimodal data. We developed a Cylindrical phase correlation method, that we aim to further improve to use on multimodal data.