Rigid Image Registration based on Pixel Grouping

We propose a pixel similarity-based algorithm enabling accurate rigid registration between single and multimodal images. The method relies on the partitioning of a reference image by a Gaussian mixture model (GMM). This partition is then projected onto the image to be registered. The main idea is that a Gaussian component in the reference image corresponds to a Gaussian component in the image to be registered. If the images are correctly registered the total distance between the corresponding components is minimum. An advantage of the proposed method is that it may handle multidimensional (vector valued) images where histogram-based methods such as the widely used mutual information is not tractable due to the high dimension of the data. Also, experimental results indicate that, even in the case of images presenting low SNR, the proposed algorithm compares favorably to the histogram-based mutual information method that is widely used in a variety of applications.