Object Detection and Localization by Dynamic Template Warping

  • Authors:

  • Aparna Lakshmi Ratan;W. Eric L. Grimson;William M. Wells, III

  • Affiliations:

  • Massachusetts Institute of Technology, Artificial Intelligence Laboratory, 545 Technology Square, Cambridge, MA 02139, USA. aparna@ai.mit.edu;Massachusetts Institute of Technology, Artificial Intelligence Laboratory, 545 Technology Square, Cambridge, MA 02139, USA. welg@ai.mit.edu;Massachusetts Institute of Technology, Artificial Intelligence Laboratory, 545 Technology Square, Cambridge, MA 02139, USA. sw@ai.mit.edu

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

Quantified Score

Hi-index 0.00

Visualization

Abstract

A simple method is presented for detecting, localizing andrecognizing instances of classes of objects, while accommodating awide variation in an object's pose. The method utilizes a smalltwo-dimensional template that is warped into an image, and convertslocalization to a one-dimensional sub-problem, with the search for amatch between image and template executed by dynamic programming.For roughly cylindrical objects (like heads), the method recoversthree of the six degrees of freedom of motion (2 translation, 1rotation), and accommodates two more degrees of freedom in the searchprocess (1 rotation, 1 translation). Experiments demonstrate thatthe method provides an efficient search strategy that outperformsnormalized correlation. This is demonstrated in the example domainof face detection and localization, and can extended to more generaldetection tasks. An additional technique recovers rough object posefrom the match results, and is used in a two stage recognitionexperiment in conjunction with maximization of mutual information.