Linear models for spectral reflectance functions over the mid-wave and long-wave infrared

We analyze the use of linear models for IR spectral reflectance functions. Linear models have been studied extensively for the visible wavelengths and form the basis of several approaches to estimating surface properties from color images. The IR analysis is performed with measured spectral reflectance functions for 394 samples of natural and man-made materials. The mid-wave (3–5 μm) and long-wave (8–12.5 μm) atmospheric windows of the IR spectrum are considered separately. Since materials tend to have stronger spectral features over the 8–12.5 μm range, linear models for the long-wave IR require more parameters than for the mid-wave IR, in order to obtain the same accuracy. We show that a six-parameter linear model provides an excellent approximation for mid-wave (3–5 μm) reflectance functions and that a nine-parameter linear model provides a satisfactory approximation for long-wave (8–12.5 μm) reflectance functions.