Systems and methods for image alignment are disclosed. A method includes: determining a first set of patches for a first image and a second set of patches for a second image, wherein each patch in the first and second sets of patches comprises a portion of the first and second image, respectively; determining a set of possible match pairings of a patch from the first set and a patch from the second set; for each match pairing in the set of possible match pairings, computing a transformation operation to align the patch from the first set of patches and the patch from the second set of patches in the match pairing; grouping the match pairings into clusters based on parameters of the transformation operations of the match pairings; and, performing the transformation operation corresponding to at least one cluster to align the first image to the second image.