Recognition without Correspondence using MultidimensionalReceptive Field Histograms

  • Authors:
  • Bernt Schiele;James L. Crowley

  • Affiliations:
  • MIT Media Laboratory, Room 384C, 20 Ames Street, Cambridge, MA 02139, USA. bernt@media.mit.edu;GRAVIR, INRIA Rhône–Alpes, 655, Avenue de l'Europe, 38300 Monbonnot, France. james.crowley@imag.fr

  • Venue:
  • International Journal of Computer Vision
  • Year:
  • 2000

Quantified Score

Hi-index 0.01

Visualization

Abstract

The appearance of an object is composed of local structure.This local structure can be described and characterized by a vectorof local features measured by local operators such as Gaussianderivatives or Gabor filters. This article presents a technique whereappearances of objects are represented by the joint statistics ofsuch local neighborhood operators. As such, this represents a newclass of appearance based techniques for computer vision. Based onjoint statistics, the paper develops techniques for theidentification of multiple objects at arbitrary positions andorientations in a cluttered scene. Experiments show that thesetechniques can identify over 100 objects in the presence of majorocclusions. Most remarkably, the techniques have low complexity andtherefore run in real-time.