The development of remote geological interpretation technology is booming during recent years.
However, there is a significant obstacle for extracting geology information from remote sensing
imagery--the presence of clouds and their shadows. Diverse techniques have been proposed including
different algorithms such as filtering algorithm and multi-temporal cloud removing algorithm to solve
the problem. This paper presents a modified solution to denoise the haze, based on ETM+ imagery.
First of all, wavelet transform is applied to Band1, Band2 and Band3 imagery to determine the clear
region and different levels of cloud regions. Then all pixels of the ETM+ imagery are classified to
specific cover types after the cluster analysis of band4, Band5 and Band7. At last, the mean reflectance
matching is performed in the first three bands separately according to different cover types in both clear
region and cloud region. Above all, the method is implemented by IDL. The results show that this
modified method not only can quantitatively determine the cloud area but also can remove cloud from
imagery efficiently. Moreover, compared with the homomorphic filtering method, the experiment
results of the proposed method is much more satisfying in Geology Interpretation.