Title :
A closed-form massively-parallel range-from-image-flow algorithm
Author :
Raviv, Daniel ; Albus, James S.
Author_Institution :
Dept. of Electr. Eng., Florida Atlantic Univ., Boca Raton, FL, USA
Abstract :
A closed-form solution for obtaining the 3-D structure of a scene for a given six-degree-of-freedom motion of the camera is provided. The solution is massively parallel, i.e., the range that corresponds to each pixel is dependent on the spatial and temporal changes in intensities of that pixel, and on the motion parameters of the camera. The measurements of the intensities are done in a priori known directions. The solution is for the general case of camera motion. The deviation is based upon representing the image in the spherical coordinate system, although a similar approach could be taken for other image domains, e.g., the planar coordinate system. Comments are made on the amount of computations, error and singular points of the solutions. A practical way to significantly reduce and implement them is suggested
Keywords :
computerised picture processing; parallel algorithms; 3D structure; camera motion; closed-form solution; image domains; massively parallel; motion parameters; range-from-image-flow algorithm; spatial changes; spherical coordinate system; temporal changes; Animals; Brightness; Cameras; Closed-form solution; Eyes; Image segmentation; Layout; Retina; Robot kinematics; Surface texture; Surface treatment;
Journal_Title :
Systems, Man and Cybernetics, IEEE Transactions on