The cloud coverage data are used to select the images for image matching in the SWIR parallax correction and the inter-telescope registration, since cloud-free images are essential for image matching. The algorithm is based on the fact that clouds have the highest reflectivity in the visible and the short wave infrared spectral region except for snow and ice, and a low emission in the thermal infrared spectral region because of their lower temperature than targets on the earth.

It will be very important to distinguish clouds from snow and ice on the earth's surface. The discrimination is carried out using knowledge that snow/ice may be brighter in band 2 and darker in band 4 than cloud.

The EOSAT algorithm is employed for cloud coverage calculation. Figure 2-11 (a) shows the algorithm flow. The bands 2, 4 and 11 data are used as representative data for the visible, short wave and thermal infrared spectral regions, respectively. Two threshold values of T2 and T2* , two threshold values of T4 and T4* and one threshold value of T11 are introduced for band 2, band 4 and band 11, respectively, as the borders between cloudy and cloud-free targets. According to this algorithm the hatched regions shown in Figure 2-11(b) are judged to be cloudy.

The calculation is carried out for every block whose size is 20 x 20 SWIR pixels. The threshold value depends on the latitude and the season.

Figure 2-11 Cloud assessment algorithm