There are some distortions in the remotely sensed imageries obtained by the airborne MSS. They are the angnlar effects caused mainly by the rough surface in the optical sence, the atmospheric effects, system distortions and other variables. CCT data for analysis inevitably contain the multiplicative noise and additional ones. However it is almost impossible to obtain theoretically these noises by using the reference data, because some of these parameters are unknown or uncertain. It is generally considered as the experimental method that the separation of the noise from signature is achieved by using the appropriate relationship between CCT data and ground surface reflectance measured exactly. The powerful reflection model, called the Equivalent Reflection Model, has been developed by us for the purpose of obtaining the precise rough surface reflectance in consideration of look angle, sun azimuth, sun zenith and other factors. This model enable us to get rid of complicated and tiresome observations. In this paper, the first order approximation formula is adopted for the relationship between CCT data and computed reflectance. As the result, we have been able to estimate the multiplicational noise coefficient and additional noise constant contained in the formula mentioned above. Moreover it has become clear that the reflectance values actually obtained from CCT data show good agreement with the computed reflectance. The multiplicational noise coefficient and additional noise constant can be estimated only by using both the CCT data and the exact ground surface reflectance. From the result described above, we may conclude that the CCT data can be converted into the reflectance by using the simple first order approximation formula relating CCT data to computed reflectance.
A methodology for purification of training samples for the pixel-wise Maximum Likelihood Classification is proposed. In this method, pixels which show comparatively high local spectral variability as well as spectrally separable classes are removed from the preliminary designated training samples. An example using agricultural TM data shows that separability can be improved 3.78 times in terms of divergence between a specific class pair, goodness of fit to Gaussian can be improved 0.14 times in terms of Chi-square, 11.9%improvement of the weighted mean percentage classification accuracy can be achieved, and most importantly a 20.6%improvement of probability of correct classification can be achieved for a specific class.