Free satellite imagery often contains a lot of clouds. Removing them by hand with the Clone Tool of an image manipulation program is a tiresome task.
- Select the clouds by color
- Grow the selection by a few pixels to eliminate pixels that belong to the cloud, but are too dark for 'select by color'
- Use the 'Script-Fu -> Enhance -> Smart Remove' tool
The clouds have gone but their shadows are still there. Roads suddenly end in a green patch. Villages that were covered by clouds are overgrown by green woods and fields. Shadows of no longer existing clouds still darken the area.
Selecting by color has the drawback that every area having a cloud-like color will be removed. Alternatively, you can select clouds by hand. The selection does not need to be accurate, but all details accidentally selected will vanish.
Professional geographers use the LANDSAT infrared bands to distinguish between clouds and white areas. This is called 'cloud detection using spectral features', which is short for: if its white/greyish in the visible bands and darkish in infrared, it's a cloud. With some refinement, clouds and snow can be separated. This might be useful for high mountain areas.
Bew is working on a tool to automatically detect clouds. You call the tool with LANDSAT images of band 1, 2, 3, 5, and 6. The tool creates a black and white image containing just the clouds as black fields. You can import this image as a layer and simply select all black fields. With this selection, call the 'Smart Remove'-tool after having changed to the layer containing your pan-sharpened LANDSAT image. The tool is not ready for public use yet.