Is anyone aware of algorithm(s) that have been developed for computing remotely sensed image footprints that handle body limbs? The limb is the edge of the body / atmosphere where only the blackness of space is then visible. Images with a visible limb are captured, possibly, when the sensor is highly oblique or at great distance from the body.
For the vast majority of remotely sensed data, the image footprint (see here for an example of Landsat footprints over texas) is roughly a square (for a framing camera) or a long rectangle (for a line scan camera).
An approach to generate a footprint for images without a visible limb can simply walk the image boundary and compute the body intersection at some node interval. For example, the upper left corner intersects the body at (x,y,z in some reference frame), the 100th pixel along the top edge intersection at (x1,y1,z1), etc. Those intersections define the footprint of the image on the body. In the case of a limb (or the terminator honestly), walking the edge is not possible because a body intersection does not exist for some subset of pixels (because we see space). A naive approach would be to create a footprint with a node at every pixel, but that could be > 70,000 geometries for some long line scan images.
Before I go building an algorithm, I wonder if one might exist.