<< Back

Geo-Registration of Aerial Images

Yaser Ajmal, Sohaib Khan

A project investigating registration of aerial images with high definition ortho-imagery. Aerial Images were first rectified to an orthographic viewing system, on the basis of a telemetry file documenting the position of the aircraft at the time of frame-capture. The orthorectified aerial image was then registered with the OrthoGraphic Reference Imagery by computing high interest matchpoints and finding correspondence. The error is calculated between the positions of the points and is minimized using the Levenberg-Marquadet Minimization Algorithm. An Update telemetry file is also generated in the process.

For Power Point presentation click here

Image Quality Metrics for Registration

Omar Javed
, Sohaib Khan

We worked on developing an image quality metric to work as a prescreener with their image registration software. The idea is to filter out images that will be hard to register before hand, so as not to waste resources during registration. Three image quality metrics: SNR estimation, Texture measures, and Gabor energy, were studied for five video clips of A. P. Hill area consisting of roughly 500 images. Using each image metric false positive, false negative, and % misclassification were computed. Then these results were analyzed. Our conclusion is that since Gabor measure is able to detect three major causes of mission imagery failure: amount of texture, cloudy images, blurry images, is more suitable than other two measures.

For Power Point presentation click here

Associated Publications

Richard W. Cannata, Mubarak Shah, Steven G. Blask, John A. Van Workum. Autonomous Video Registration Using Sensor Model Parameter Adjustments , Applied Imagery Pattern Recognition Workshop (AIPR) 2000, Cosmos Club, Washington D.C., Oct 16-18, 2000

Book Chapters


Chapter in Video Registration

Y. Sheikh, S. Khan, M. Shah and R. Cannata
Mubarak Shah and Rakesh Kumar
Kluwer Academic Publishers, 2003


Workshop on Video Registration
With International Conference on Computer Vision, ICCV 2001