Entire Building Shape Recovery from Near Distance Images.
Tsuyoshi Migita, Akira Amano, Naoki Asada · Machine Vision Applications · 2002
This paper describes a method to recover entire 3D shape of a building from an image sequence. This is a kind of Shape and Motion problem, whereas conventional methods do not work well with images taken around a large object in near distance. Since such image sequence is a set of partial observations, 3D recovery becomes unstable. We first discuss the property of local minima of the nonlinear optimization function, and then describe a procedure to find the global minimum by avoiding pseudo solutions. Experiments using real images have shown that the proposed method successfully recovered 3D shapes from eleven sets of image sequence.