The digital counts or the scanned densities of the Landsat MSS image were

transformed into absolute radiance values. Each radiance of the four bands were decomposed into three parts with different significances and then the parts with the similar significance from the four bands were recombined. As a result

we obtained three new variables. Which are named respectively the general radiance level （L）

the visible-infrared （radiance） balance （B） and the band radiance) variation vector （V）. The L is powerful in monitoring radiation intensities of ground features; the B is powerful in mainly monitoring the water

the moisture or wetness of ground features and in differentiating the cloud

snow and ice among them; the V is powerful in monitoring density of plant cover and its vigor and hence is an excellent indicator of vegetation conditions. Those three variables contain nearly all the information of the original four bands

so we need only manipulating the three instead of four variables. The algorithm is based on the regression analysis and is very simple. Lower dimension and simpler calculation will lead to time and monev saving.Using this method

we can obtain "real" （not false） color composites

which are real three-coloured and near to the natural color. And we can obtain significant

corresponding to the ground feature characteristics

density slicing images.The methodology described above was applied to the data on hand from China and Canada and other data found in literatures from Asia

Europe

Africa

Australia

Oceania and North and South America.