Denoising And Cloud Shadow Removing Of Remote Sensing Image Based On Contourlet Transform

As a new image description approach in multi-scale geometric analysis, contourlet transform, with its excellent characteristics of multi-direction and anisotropy, is a real description approach for 2D image. And curves in image can be represented sparsely with contourlet transform. Compare with wavelet transform, contourlet transform can seize the line singularity of image more effectively and express the texture information of the image's edges more clearly. Researches and experiments on the aspects of denoising in infrared remote sensing image and cloud shadows removing in color remote sensing image were executed. Detail research contents and experiment results were as follow:1. With the introduction of wavelet transform theory and its disadvantage in image processing, multi-scale geometric analysis approach was recommended. While the Basic principles and structure of contourlet transform were introduced in detail.2. Based on the combination of contourlet transform and Bayesian bivariate model, an algorithm for infrared remote sensing image denoising was proposed. With contourlet transform's characteristics of multi-direction and anisotropy, contourlet coefficients of noise image were acquired. And Bayesian bivariate model was used to exploit the dependencies of the contourlet coefficients across the scales. Simulation results demonstrated that both PSNR and visual effects of the image processed by the proposed algorithm were improved remarkably, especially for those images rich in texture detail.3. Based on contourlet transform, an algorithm for cloud shadow removing in color remote sensing image was proposed. It consisted of shadow region detection and shadow removal. Firstly, shadow region detection based on contourlet transform was introduced; then, compensation of both low and high frequency for the shadow area was executed; finally, we performed transformation of color space between HSV and RGB on the compensated image and got the image after cloud shadow removal. Experiment results demonstrated that the proposed method, compared with the wavelet transform, could remove cloud shadow in the color remote sensing image more effectively, and recovered the surface features information in the shadow better.