Overlapping fields with image processing: Image stitching

Image stitching or photo stitching is the process of combining multiple photographic images with overlapping fields of view to produce a segmented panorama or high-resolution image. Commonly performed through the use of computer software, most approaches to image stitching require nearly exact overlaps between images and identical exposures to produce seamless results, although some stitching algorithms actually benefit from differently exposed images by doing high-dynamic-range-imaging in regions of overlap. Some digital cameras can stitch their photos internally. nce the illumination in two views cannot be guaranteed to be the same, stitching two images could create a visible seam. Other reasons for the seam appearing could be the background changing between two images for the same continuous foreground. In general, the major issues to deal with are presence of parallax, lens distortion, scene motion, and exposure differences. For panoramic stitching, the ideal set of images will have a reasonable amount of overlap (at least 15–30%) to overcome lens distortion and have enough detectable features. The set of images will have consistent exposure between frames to minimize the probability of seams occurring. In the non-ideal real-life case, the intensity varies across the whole scene and so does the contrast and intensity across the frames. Lens distortion, motion in the scene, and misalignment all cause ghosting. Also, the ratio of width to height of a panorama image needs to be taken into account to create a visually pleasing composite.