LANDSAT MSS installes six different detectors per band, which enable a simultaneous scan of six lines at once. However, a cyclic pattern of line noise can be seen in the LANDSAT imagery, because of small differences among six different detectors. This pattern is apparent especially in the area of water or forest, where reflectance is uniform and low. The paper describes a methodology to correct radiometric differences among those detectors, by using accumulated histogram. Based upon the fact that respective accumulated histograms for each detector will become equal statistically, a methodology was proposed to correct original data to make almost same distribution within a certain allowance for six accumulated histograms corresponding to six different detectors.
Mean and Standard Deviation Equalization Method (MSDEM) generally used in GSFC, CCRS, or NASDA to make the radiometric adjustment between 6 detectors of each Landsat MSS Band, can not always produce the correct imageries. The uncorrect results happen in the scenes covered with a fairly plenty of clouds. Investigating the histogram of such imageries, the authors reach a conclusion that the MSDEM method needs an assumption that the scene should include no pixels overflown radiometrically. Level Matching Method using Cumulative Frequency Distribution (LMMCFD) satisfies any case of scenaries including overflown pixels. In Figure 2, (e) and (f) are the results of LMMCFD method from the raw imageries (a) and (b), while unsufficient correction remains in (c) and (d) processed from the same raw data by the MSDEM method of NASDA.
Vegetation reflectance plays as a noise source in extracting soil information from vegetated area. In this paper, a simple optical model which explained the influence of vegetation reflectance on soil reflectance is proposed and the soil index which represents soil types independent of vegetation cover, is described. This soil index is considered to be sensitive to the degree of the soil organic matter content. The usefulness of this soil index was investigated and tested by the channel selection based on the statistical distances among soil classes and by multispectral apttern recognition. As a result, by combining the SI-value with other 3-channel which represent vegetation reflectances, various types of soil with 0-50 % vegetation covers are accurately classified.