At the core of the i2k Align™ technology is the powerful i2Align™ image registration algorithm, which both decides whether or not two images overlap and computes the inter-image transformation (pixel location mapping function) between them. Hence it is both a registration and recognition algorithm. In i2k Align version 2.0, this is combined with a new algorithm to compute the topological arrangement of all the images, even when there are ambiguities in the image set. Once the arrangement is decided upon, i2k Align computes the transformations that combine all the images, mapping them into a common "aligned" coordinate system, by employing an "unbiased" estimation technique that produces accurately-aligned images with minimal geometric distortion despite low inter-image overlaps.
In order to produce accurately-aligned and seamlessly-montaged images, much more is needed than just i2Align™ and our topological arrangement algorithm. DualAlign™ has developed a host of supporting technologies that work together to make the i2k Align™ software powerful, widely-applicable, and easy to use. These technologies include automatic, masking, single and multi-image vignetting correction, illumination modeling and camera-parameter estimation, inter-image seam selection and final composite blending. They are used in different combinations for different choices of tools and image types to produce the results. For example, montaging of photographic images requires all technologies except perhaps for masking in order to produce the final, seamless result. Montaging thermal images, on the other hand, does not include vignetting and camera estimation in order to preserve the recorded intensities, but does include seam selection.