Local Computation of Shape

Pentland, Alex

Local Computation of Shape

AAAI 1982 pp. 22-25

/aaai/1982/pentland1982aaai-local/

Abstract

Local, parallel analysis of shading in an image may be used to infer surface shape without o priori knowledge about the viewed scene. The following has been proven: Given a reasonable assumption about surface smoothness, local analysis of the second derivative of image intensity allows the image-plane component of surface orientation to be determined exactly, and a maximum-likelihood estimate of the remaining depth component of orientation to be made. An algorithm has been developed to compute surface orientation, in parallel, for each image point without knowledge of scene characteristics. This algorithm has been evaluated on both natural and synthesized images, and produces a good estimate of shape.

PDF AAAI Semantic Scholar

Cite

Text

Pentland. "Local Computation of Shape." AAAI Conference on Artificial Intelligence, 1982.

Markdown

[Pentland. "Local Computation of Shape." AAAI Conference on Artificial Intelligence, 1982.](https://mlanthology.org/aaai/1982/pentland1982aaai-local/)

BibTeX

@inproceedings{pentland1982aaai-local,
  title     = {{Local Computation of Shape}},
  author    = {Pentland, Alex},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year      = {1982},
  pages     = {22-25},
  url       = {https://mlanthology.org/aaai/1982/pentland1982aaai-local/}
}