Cu = NY = 
f 
n 
8 
f 
Is 
0 
at 
Ali Farrag, Farrag 
  
desert areas and paved roads. Such common types of land cover were used to compare the reflectancy (DNs) in the 
green band of 1995 image and 1998 image compared with their reflectancy in the same band of 1992 image. Then the 
green band of images of 1995 and 1998 was shifted by certain value to be comparable with the reflectancy in the same 
band of image 1992. It is found that the green band of 1995 image need to be shifted by +5 while the green band of 
1998 image need to be shifted by +30. The histograms of the adjusted green bands are given in Figure 2. One can realize 
that there is some similarity between the histograms (especially with respect to the pixels that represented by small 
DNs). The little shift can be due to difference in crop’s age because the image of 1992 and the image of 1995 were 
captured on June while the image of 1998 was captured on August. 
  
Frequency 
  
  
"d 7 ame ee S v > Un T T T p E 
55 80 105 130 155 180 205 230 255 
Digital Number (DN) 
| Green band (92) - - Green band (95) -- Green band (98) | 
  
  
Figure 2. Histograms of green bands after adjustment of DNs 
b- Stretching and slicing image bands to a number of levels 
The process of stretching and slicing image bans to a number of levels was carried out on a smaller portion of the image 
that corresponding to 12 km x 17 km on the ground as identified in step 1 above. The process of linear stretching 
involves identification of two brightness values (DNs) as minimum and maximum values in the resulting stretched 
image. These minimum and maximum values represent the limits of the feature of interest (the polluted area in our case). 
Then the image data between the two limits was sliced to a number of levels, to be interpreted as pollution levels. 
As a result of examining the image of each band, it found that, the polluted pixels are located at the extreme end of the 
green and the red bands (have high digital numbers) and near the end at the near infra red band. Also these pixels are 
highly separable from the surrounding pixels in all bands [Farrag (1997)]. As mentioned above, the analyzing of 
information contents of the green band of SPOT images is emphasized, in this study. After several tests on the green 
band of 1992 image it has been found that the minimum brightness value of the threshold to extract the polluted pixels 
has been defined to be 155 [Farrag (1997)]. The same value can be used for stretching the green bands of 1995 image 
and of 1998 image, because the DNs of these image bands were adjusted to be comparable with those of the green band 
of 1992 image. All pixels with spectral response above that threshold (within the specified window) are corresponding 
to polluted areas, while, all pixels with a spectral response below that threshold are corresponding to other classes. 
Reasonable results were obtained by selecting six levels in the stretched green band of the two images. The resulting 
images of streatched green band of 1995 image and 1998 image are given in Plates 5 and 6 respectively. 
The six classes of the stretched green bands can be identified as: 
- One class of zero brightness corresponds to the background (uncontaminated area). 
- Four classes of brightness variations corresponding to polluted areas, with different degrees of intensities. 
- One class of brightness corresponding to the open pit mine (to the west of the factory) and some places within the 
factory. This class is identified as high contaminated areas. 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 49