Recovering shape and motion from undersea images

Work on the development of passive-vision sensing capabilities that can increase the autonomy of an undersea vehicle is reported. Two classic machine-vision problems for recovering motion and the shape of objects from their images are considered, in light of formulations that incorporate effects due to the illumination attenuation and nonuniform lighting commonly encountered in undersea environments. Based on the formulations, methods are proposed for recovering the orientation of a planar patch from its image shading and for estimating the image motion, a useful tool in dynamic scene analysis