Registering Oblique Image Pairs Without Camera Models

Abstract

This paper presents automatic electro-optical image registration of images taken from oblique views (large baseline separation). The method uses feature based point correspondence to solve for the projective parameters in an iterative approach. The feature points are solved for by analyzing region contours on segmented images. A simple geometric invariance method is used to match points and remove outliers. A projective transform is applied to previously warped images until the tie point correspondence satisfies a suitable error measure. The series of projective transformation parameters are then combined to solve for the overall global transformation.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
May 24, 1999
Accession Number
ADA392036

Entities

People

  • Robert M. Taylor

Organizations

  • Johns Hopkins University

Tags

Communities of Interest

  • Air Platforms
  • Materials and Manufacturing Processes
  • Sensors

DTIC Thesaurus Topics

  • Aerial Photographs
  • Algorithms
  • Boundaries
  • Cameras
  • Change Detection
  • Computer Vision
  • Correlation Techniques
  • Detection
  • Detectors
  • Feature Extraction
  • Frequency
  • Image Processing
  • Image Registration
  • Images
  • Optical Images
  • Physics Laboratories
  • Two Dimensional

Fields of Study

  • Physics

Readers

  • Electromagnetic Wave Scattering and Antenna Radiation Engineering
  • Human-Computer Interaction (HCI).
  • Mathematical Modeling and Probability Theory.