Research On Removing Thin Cloud In Single Remote Sensing Image

In the actual application of remote sensing image, there are inevitably situations that some local image covered by clouds, and we can improve the efficiency of the utilization rate of image data and the accuracy of the interpretation if the clouds are removed at the image preprocessing stage. Most of the methods in removing the clouds noise at the remote sensing images now mainly aimed at the clouds which are in wide range of low frequency part, and without enough attention for the processing of the clouds noise of high frequency local mutations. This paper, after studies of the basic theory, the model of cloud and common methods of cloud removing, in view of the single remote sensing image which is without the ancillary information, puts forward an improved method based on the wavelet transform which is aimed at removing the clouds noise at the high frequency local mutations. The main research work included:1. Study the theory of Fourier transform and wavelet transformation further, analysis the model of clouds in the remote sensing image from the formation of clouds principle, and study and summarized the commonly used methods for removing clouds noise which are including: homomorphism filter method, Laplace airspace enhance method and Traditional wavelet variation method;2. Specifically against single shot remote sensing image, researching and summing up the advantages and disadvantages of above methods, in view of weakness of common methods which can not remove the clouds noise of the high frequency local mutations area, combined with the classical methods of the airspace image processing and frequency domain image processing, put forward a new method which is combined the ideas of the median filter and wavelet transform;3. In specific theory is conducted on the basis of experiment, comparing the effect of traditional wavelet transformation and improve wavelet transform, and used the contrast test such as:homomorphism filter and Laplace enhancement, analysis shortcomings of the new method;4. From the subjective visual evaluation and objective statistical index to evaluate the effect of above methods, searching for the improving based on the results of the evaluation, give the future research direction.