Date of Award

Spring 5-2011

Degree Type


Degree Name

Doctor of Philosophy (PhD)




Computing Sciences and Computer Engineering

Committee Chair

Benjamin Seyfarth

Committee Chair Department


Committee Member 2

Dia Ali

Committee Member 2 Department


Committee Member 3

Beddhu Murali

Committee Member 3 Department


Committee Member 4

Ras Pandey

Committee Member 4 Department

Physics and Astronomy

Committee Member 5

Chaoyang Zhang

Committee Member 5 Department



The process of creating aerial photo mosaics can be severely affected by clouds and the shadows they create. In the CZMIL project discussed in this work, the aerial survey aircraft flies below the clouds, but the shadows cast from clouds above the aircraft cause the resultant mosaic image to have sub-optimal results. Large intensity variations, caused both from the cloud shadow within a single image and the juxtaposition of areas of cloud shadow and no cloud shadow during the image stitching process, create an image that may not be as useful to the concerned research scientist. Ideally, we would like to be able to detect such distortions and correct for them, effectively removing the effects of the cloud shadow from the mosaic.

In this work, we present a method for identifying areas of cloud shadow within the image mosaic process, using supervised classification methods, and subsequently correcting these areas via several image matching and color correction techniques. Although the available data contained many extreme circumstances, we show that, in general, our decision to use LIDAR reflectance images to correctly classify cloud and not cloud pixels has been very successful, and is the fundamental basis for any color correction used to remove the cloud shadows. We also implement and discuss several color transformation methods which are used to correct the cloud shadow covered pixels, with the goal of producing a mosaic image which is free from cloud shadow effects.