Research On Cloud Removal Of Methods Of Remote Sensing Image Based On Multi-resolution Analysis

With the development of technology of remote sensing and aerospace, remote sensing image has been applied in a wide range of applications. Remote sensing image is affected by the factor of weather easily. Therefore there are cloud-covering areas in the image, which affect the further procession and recognition of the image, but also reduce the utilization rate of images. So it is very meaningful to study the effective methods to remove cloud of remote sensing images.This paper analyzes the cause of the cloud formation, the theory of multi-resolution analysis and its advantage in the field of image processing, then thin and thick cloud removal methods based on multi-resolution analysis are studied. The main works and achievements are as follows:1. The theory and implementation of wavelet transform and the multi-resolution analysis was introduced. The development overview of multi-resolution analysis of image and the implication of multi-scale geometrical analysis in the cloud removal was discussed.2. A new method based on curvelet transform to remove thin cloud of remote sensing image was proposed. Firstly, the source image with thin cloud was decomposed using curvelet transform. Then cloud detection was applied in the low frequency coefficients and the cloud covering area was compensated, whereas the high frequency coefficients were compensated by adaptive threshold enhancement. Finally the coefficients were reconstructed by curvelet inverse transform to retain the cloud removal result. Experimental results show that with the comparison of homomorphic filtering method and wavelet transform, this method can achieve better effect of thin cloud removal, moreover it can preserves the image detail and edge information effectively.3. Taking advantage of the nonsubsumpled contourlet transform which is a fully shft-invariant, multiscale, and multidirection, an algorithm based on nonsubsumpled contourlet transform to remove thin cloud of remote sensing image was proposed. Firstly, the source image with thin cloud was decomposed using nonsubsumpled contourlet transform. Then all the subband coefficients were compensated. Finally the coefficients were reconstructed by nonsubsumpled contourlet transform to retain the cloud removal result. The experimental results indicate that this algorithm can get better cloud removal effect and have more clear texture information than others.4. The algorithm based on support vector machine is presented to remove thick clouds in remote sensing images. Firstly support vector value contourlet transform is constructed by the theory of support vector machine and the images are decomposed. Then cloud detection and image fusion by different fusion rules are carried out. Finally the cloud removal images are got by reconstructed images. For the remote sensing images with the thick clouds and no overlapping cloud area, the experimental results show that the algorithm can achieve better result and the result images are clearer.