Title :
Accuracy of the regularised dynamic programming stereo
Author :
Gimel´farb, Georgy ; Lipowezky, Uri
Author_Institution :
Dept. of Comput. Sci., Auckland Univ., New Zealand
Abstract :
Binocular stereo is an ill-posed problem that has to be regularised with respect to partial occlusions and homogeneous textures for obtaining a unique solution. We analyse typical errors of the dynamic programming stereo based on a probabilistic regularisation of partial occlusions. Experiments show that the reconstruction is accurate until occlusions do not prevail over the binocularly visible terrain.
Keywords :
Markov processes; dynamic programming; image texture; probability; stereo image processing; binocular stereo; homogeneous textures; ill-posed problem; partial occlusions; probabilistic regularisation; regularised dynamic programming stereo; unique solution; Computer science; Distortion; Dynamic programming; Error analysis; History; Image reconstruction; Pixel; Stereo image processing; Surface texture; Systems engineering and theory;
Conference_Titel :
Pattern Recognition, 2002. Proceedings. 16th International Conference on
Print_ISBN :
0-7695-1695-X
DOI :
10.1109/ICPR.2002.1048015