The NASA Ames Stereo Pipeline is a suite of automated geodesy & stereogrammetry tools designed for processing planetary imagery captured from orbiting and landed robotic explorers on other planets.
Our distortion models, all of them, even the most basic ones, suffer from a big issue. We assume that distortion is little, and that distorted and undistorted pixels are basically in the image bounding box, plus or minus 10-100 pixels or so. That is true for Photometrix and most frame camera with minimal distortion. It is not true at all of FishEye cameras. It may not be true for KH-4B. Undistortion transforms the image box into a non-linear region, that can be partially in the image box and partially outside of it. We should at least keep track of thee bounding box of the domain of undistortion. When the distortion parameters are floated, this region changes. When we convert a pinhole model, we likely don't sample this region correctly.
oleg-alexandrov
commented
10 months ago
Our distortion models, all of them, even the most basic ones, suffer from a big issue. We assume that distortion is little, and that distorted and undistorted pixels are basically in the image bounding box, plus or minus 10-100 pixels or so. That is true for Photometrix and most frame camera with minimal distortion. It is not true at all of FishEye cameras. It may not be true for KH-4B. Undistortion transforms the image box into a non-linear region, that can be partially in the image box and partially outside of it. We should at least keep track of thee bounding box of the domain of undistortion. When the distortion parameters are floated, this region changes. When we convert a pinhole model, we likely don't sample this region correctly.