An ensemble of terrain images from forested and semiarid environments are analyzed in terms of first- and second-order textural statistics and Fourier and wavelet transform metrics. Such parameters are sought in an effort to reduce the dimensionality of terrain image information to suitable levels of both generality and specificity. By developing a practical set of feature metrics, a real and generated scene can be compared critically in terms of scene elements rather than pixel-to-pixel error. This paper presents some of the results of such a validation process for the Smart Weapons Operability Enhancement (SWOE) Joint Test and Evaluation (JT&E) program. Statistical and transform-based techniques were applied to images obtained from Grayling, Michigan, and Yuma, Arizona, at various times of day under a variety of weather conditions. Statistical analyses of scene radiance distributions and "clutter" content were performed both spatially and temporally. An emphasis on the spatial and temporal distinction between widely distributed terrain features (grass, dirt) and discrete features (trees, bushes) is made.<<ETX>>
[1]
Walter I. Futterman,et al.
Estimation of scene correlation lengths
,
1991,
Defense, Security, and Sensing.
[2]
J. Dubois,et al.
Evaluation Of The Grey-level Co-occurrence Matrix Method For Land-cover Classification Using Spot Imagery
,
1990
.
[3]
N N Ben-Yosef,et al.
Natural terrain infrared radiance statistics: daily variation.
,
1985,
Applied optics.
[4]
Y. Itakura,et al.
Statistical properties of the background noise for the atmospheric windows in the intermediate infrared region
,
1974
.
[5]
Leslie S. Balfour,et al.
Different aspects of backgrounds in various spectral bands
,
1991,
Defense, Security, and Sensing.
[6]
N Ben-Yosef,et al.
Temporal behavior of thermal images.
,
1985,
Applied optics.
[7]
N. Ben-yosef,et al.
Measurement and analysis of 2-D infrared natural background.
,
1985,
Applied optics.