Space-time Interest Points

  • Authors:
  • Ivan Laptev;Tony Lindeberg

  • Affiliations:
  • -;-

  • Venue:
  • ICCV '03 Proceedings of the Ninth IEEE International Conference on Computer Vision - Volume 2
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

Local image features or interest points provide compact andabstract representations of patterns in an image. In this paper, wepropose to extend the notion of spatial interest points into thespatio-temporal domain and show how the resulting features oftenreflect interesting events that can be used for a compactrepresentation of video data as well as for its interpretation. Todetect spatio-temporal events, we build on the idea of the Harrisand Förstner interest point operators and detect localstructures in space-time where the image values have significantlocal variations in both space and time. We then estimate thespatio-temporal extents of the detected events and compute theirscale-invariant spatio-temporal descriptors. Using suchdescriptors, we classify events and construct video representationin terms of labeled space-time points. For the problem of humanmotion analysis, we illustrate how the proposed method allows fordetection of walking people in scenes with occlusions and dynamicback-grounds.