The Interactive Planetary Image-Processing System

Segmentation of clouds using cluster analysis

“…Three infrared threshold values of 100, 135 and 188 from a grey scale of 0-255 were chosen interactively from figure 2 to represent the transitions between the four categories. The thresholding algorithm (Hunt et al 1984) then modifies these thresholds to take into account the available solar energy, based on the time, day and month of the image and on the visible image radiance values. This results in the image of figure 3 where brighter shades of grey represent higher cloud and black represents no cloud.…”

“…Visible (0,4-1,1 j.lm)and infrared, water-vapour absorption (5,7-7,1 j.lm) images were also used, giving a three-dimensional representation of the data. The analysis is performed using the image-processing facilities of the Centre for Remote Sensing at Imperial College (Hunt et al 1984) which limits the size of the images being viewed to 512 x 512 pixels. Hence, as much of the accuracy assessment of the procedure is visual, the study area is analysed in four sections, rather than subsampling the area to fit the hardware requirements.…”