The five imagers (and lenses) of RedEdge are not registered or aligned mechanically. This is because the level of precision to be able to do this mechanically and get good alignment results is quite high (and unrealistic). In addition, shock, vibration, or handling of the camera can easily shift the alignment slightly, enough to be very noticeable in the output image.
Alignment of the bands is handled in post-processing of the data. Many third party post-processing applications (like Pix4DMapper and Agisoft for example) have workflows for creating aligned orthomosaics.
Can I align the images myself?
For customers that desire to create their own processing workflows, it is possible to align RedEdge images, but this requires some experience in computer vision or image processing. The alignment is a function of the distance of the target, so single-image alignment may not work well for images that are taken within 10 meters of objects, such as the ground or trees. It's also important to correct the lens distortion in the images before attempting to align. Correcting the lens distortion ensures that straight lines in the images are actually straight. Without correcting for lens distortion, alignment is very difficult.
For those that want to create their own alignment workflows, images from RedEdge are stored in the open and commonly supported TIFF image format, and metadata is stored in the Exif format.
We also publish processing tutorials on our github. The README contains all of the details:
Third party libraries for image alignment
There are a number of third party libraries that may provide software which will help align images, such as OpenCV or the Matlab image processing toolbox. While we don't currently provide custom support to users that are developing their processing workflows, we're happy to answer questions that might come up about the information stored in RedEdge images to help customers get the most from their RedEdge purchase.
While we can't provide direct support for these methods, each of these software have a large and active community that may be able to help you through the challenges of working with their libraries.
See the ZIP file with 5 image files attached to this article for an example of what a typical single capture will look like as far as alignment of the bands is concerned.