APPLYING BAND RATIOS TO HYPERSPECTRAL IMAGERY DATA

Most common image enhancement techniques have successfully made the transition from multispectral to hyperspectral data. These image processing techniques range from simple stretching methods to enhance the dynamic range of the imagery data to more complex basis rotation methods such as Principal Component Analysis (PCA) and variants thereof. One method that has not met with wide acceptance in the realm of hyperspectral imagery is band ratioing, i.e., dividing one band by another band. The most common reasons for not using band ratios appear to be related to sloping spectra and atmospheric heterogeneities.

Band ratioing has a long history of successful application to multispectral data. Band ratios are used to suppress illumination differences attributable to surface albedo, look angle, and topographic effects. As pointed out by Vincent (1997), and Crippen (1988) proper data preparation is essential for the successful use of band ratios. There are two spectral components that must be accounted for when using band ratios: additive and multiplicative. The additive (sensor calibration and path radiance) component must be removed prior to performing a band ratio. The additive component to the signal can be most easily removed by performing a dark subtraction on each band. The multiplicative (radiance illumination and atmospheric absorption) component must be nearly equal for each band in the ratio. The best method for making the multiplicative component nearly equal for each band in the ratio is to choose contiguous, or nearly contiguous bands. Further quantitative support for the use of band ratios is given based on derivatives of spectra. Using derivatives constraints on the distance in wavelengths between bands used for ratios.

It is also important to remember that band ratios work best when applied to distinct absorption features such as the kaolinite (2135 – 2225 nm) doublet versus broad Fe absorption features located at wavelengths less than 1000 nm.

Band ratios are an excellent tool for performing preliminary evaluations of hyperspectral data. Examples are given for Copper Flat porphyry copper deposit and Cuprite, NV to demonstrate band ratio’s capabilities as a reconnaissance tool.