Feature saliency from noise variations in invariants

Mark Jenkinson, Michael Brady

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review


Object locatisation and recognition are fundamental problems in computer vision. The goal is to enable a mobile robot to locate general, non-polyhedral objects in complex settings. This requires considerable robustness and reliability, and so low level invariants are used as a robust starting point. In particular, a set of quantities is developed that are both geometrically and photometrically invariant. They are arranged as the components of a description vector, which are matched to locate model instances. The paper analyses the variations of the invariant quantities. These arise in practice due to image noise and spatial quantisation: the case of image noise is treated here, quantisation being the subject of ongoing work. The noise models obtained show good agreement with experimental results. A probability model for the variation of the description vectors is derived and used to define a saliency measure in the image. Combining this with a Non-Uniform selection strategy in a modified RANSAC (NU-RANSAC) scheme leads to a dramatic improvement in the probability of correctly matching points, which is the basis of localising the desired object.

Original languageEnglish
Title of host publicationComputer Vision - ACCV 1998 - 3rd Asian Conference on Computer Vision, Proceedings
EditorsRoland Chin, Ting-Chuen Pong
PublisherSpringer Verlag
Number of pages8
ISBN (Print)3540639314, 9783540639312
Publication statusPublished or Issued - 1 Jan 1997
Externally publishedYes
Event3rd Asian Conference on Computer Vision, ACCV 1998 - Hong Kong, Hong Kong
Duration: 8 Jan 199810 Jan 1998

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Other3rd Asian Conference on Computer Vision, ACCV 1998
CountryHong Kong
CityHong Kong

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this