|Image alignment/registration software
|Image Registration and Fusion Systems
Analysis of two or more images of a scene often requires spatial alignment of the images so that corresponding scene
points in the images will have the same coordinates. This is required when creating a mosaic from two or more
overlapping images, finding scene changes in multitemporal images, combining information in multimodality images,
and separating moving objects from the background in a video obtained by moving platform. Following are example
applications of image alignment.
Image mosaicking: By aligning images (a) and (b) at their overlap area, a larger images is created as shown in
(c), representing the mosaic image. For more examples of image mosaicking, see the image mosaicking page.
Change detection: By aligning images of a scene taken at different times, it becomes possible to find changes
occurring in the scene over time. This capability can be used to compare before and after disaster images and assess
damages caused by a natural disaster. Images (a) and (b) show a segment of a New Jersey shore taken before and
after a hurricane. These images are courtesy of NOAA. Image (c) shows alignment of images (a) and (b). Image (a) is
shown in light blue while image (b) is shown in red. Red and light blue regions identify areas of change in the scene.
Image fusion: By aligning multimodality images, information in different modalities can be combined to an image
containing a richer source of image. The example below shows alignment of an optical image (a) and a thermal image
(b) to create a combined image (c) that contain information about both optical and thermal images. Images (a) and (b)
are courtesy of USGS.
Foreground/background separation: When images are captured from a moving platform, image motion can
be due to camera motion, object motion, or both. To separate object morion from camera motion, images are
registered at the background. Since background is static, once video frames are aligned at the background, only
motion due to the objects remain, which becomes relatively easy to detect. Images (a) and (b) show two video frames
taken moments apart by a moving platform. Registration of the images is shown in (c). The green and blue bands in
image (c) are the same as the green and blue bands in image (a), and the red band in image (c) is the same as the
red band in image (b) after it is transformed to align with image (a). The combined image shows the static background
in the natural color of the scene, while moving objects (cars and tree branches) in the scene appear as red or light
blue regions. The light blue cars in the bottom-left part of image (c) show cars in image (a) that have moved out of
view in image (b).