Use of polynomial transformations for registration of airborne digital line scan images
Abstract
Accurate image to image registration is described as crucial for multitemporal image analysis. It is noted that registration of airborne line scan images is hampered by platform instability and viewing perspective. The degree to which polynomial transforms may be used in registering two airborne line scan images is investigated. Two thermal infrared line scan images obtained over the same flight line are used as test images. Polynomials of varying numbers of terms (a maximum of 29) are derived and accuracy is tested employing various numbers of control points to determine the polynomial coefficients. The best expected image to image registration using polynomial transform methods is 1.0 pixel (RMS error) in the along-track direction and 0.75 pixels (RMS error) in the cross-track direction. Sufficient accuracy is obtained by six term polynomials using approximately 25 control points.
- Publication:
-
In: Canadian Symposium on Remote Sensing
- Pub Date:
- 1981
- Bibcode:
- 1981csrc.proc..635L
- Keywords:
-
- Error Analysis;
- Image Processing;
- Infrared Imagery;
- Infrared Scanners;
- Polynomials;
- Remote Sensing;
- Aerial Photography;
- Angular Correlation;
- Least Squares Method;
- Photomapping;
- Root-Mean-Square Errors;
- Topography;
- Instrumentation and Photography