Translator Disclaimer
10 September 2005 Data fusion: a consideration of metrics and the implications for polarimetric imagery
Author Affiliations +
Abstract
The increasing availability of multispectral, hyperspectral, and multisensor imagery during the past decade has motivated rapid growth in image fusion research for remote sensing application. While it is generally the goal of image fusion methods to obtain more information fiom the combination of multiple images than could be obtained from individual images, the measure of how well fused images actually achieve this goal is still largely subjective. Furthermore, in the selection of image data, and the analytical procedures to process this data, we make a sequence of implicit assumptions that need to be reviewed. Metrics are used to specify image characteristics necessary to perform specified tasks. New metrics are necessary to characterize the performance of image fusion techniques and also to determine the extent to which these techniques may provide more useful information than could be derived fiom non-fused imagery. Without metics, we cannot predict what data we need, or how to collect and to analyze it. There is currently no metric that encompasses both spatial and spectral resolution characteristics. A metric describing the quality of polarimetric imagery is an example of the larger problem of metrics required to specify the necessary characteristics of fused, multidimensional image data. Since polarimetric imagery is based upon the differences of image pairs obtained with the polarizer oriented orthogonally about the optic axis, misregistration introduces a false clutter that degrades information content of polarimetric imagery, so that a polarimetric image characteristic will depend upon registration accuracy. A General Image Quality Equation (GIQE) is a multivariate regression of the image quality metric against the independent imaging ammeters, such as registration in the case of polarimetric imagery. We need a General Image Quality Equation (GIQE) for polarimetric images in which one regression term describes misregistration. We need image quality metrics for polarimetric, multi-dimensional and fused multidimensional image data. In this paper we shall consider what metrics are needed to design and collect data, what general principles should guide the collection of analysis of data, and we shall consider polarimetric imagery as a simple example of a type of image fusion and analysis.
© (2005) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Michael J. Duggin and Mark L. Pugh "Data fusion: a consideration of metrics and the implications for polarimetric imagery", Proc. SPIE 5888, Polarization Science and Remote Sensing II, 588813 (10 September 2005); https://doi.org/10.1117/12.614553
PROCEEDINGS
11 PAGES


SHARE
Advertisement
Advertisement
RELATED CONTENT

Present status and trends of image fusion
Proceedings of SPIE (October 29 2009)
Pixel-registered image fusion
Proceedings of SPIE (July 04 1995)
Automatic registration between SPOT and TM images
Proceedings of SPIE (September 24 1998)

Back to Top