Title :
Likely local shape
Author_Institution :
Dept. of Comput. Sci. & Eng., California Univ., San Diego, CA, USA
Abstract :
A standard probabilistic formulation of a local shape estimator due to Witkin (1980) has been demonstrated to work to some extent on a range of data, but it also has been shown to overestimate slant. The author identifies and analyzes three sources of this error: the choice of prior probability distribution on shape; a systematic bias in the observation of image events; and a kind of nonergodicity in the stochastic process associated with surface events. As a result of this analysis, the author formulates a corrected probabilistic local shape estimator, proves a correctness theorem for it, and demonstrates its superiority to the Wilkin estimator in a set of experiments
Keywords :
pattern recognition; probability; image event observation; local shape estimator; nonergodicity; prior probability distribution; standard probabilistic formulation; stochastic process; surface events; systematic bias; Image analysis; Needles; Optical imaging; Optical noise; Probability distribution; Random variables; Shape; Stochastic processes; Surface treatment; Yield estimation;
Conference_Titel :
Computer Vision and Pattern Recognition, 1989. Proceedings CVPR '89., IEEE Computer Society Conference on
Conference_Location :
San Diego, CA
Print_ISBN :
0-8186-1952-x
DOI :
10.1109/CVPR.1989.37897
