Patent attributes
Methods for stitching multiple sub-images together to form a substantially seamless composite image are disclosed. Overlap regions formed by each pair of neighboring sub-images are periodically examined and key features common to the overlap regions in each sub-image of the pair are identified. A transformation is determined for each sub-image pair based on the positions of these key features. The transformation is split between the sub-images and applied to distort the overlap regions in each sub-image pair such that they are substantially aligned. Applying the transformations to each overlap region in the overall image enables creation of a substantially seamless composite image. In some embodiments, the process wherein the transformations are determined is run as a feedback loop to enable continuing refinement of the transformations.