SAR Image Processing Based On Wavelet-Contourlet Transformation

Wavelet based contourlet transform is a new kind of image representation which is multi-resolution, local supporting and multi-direction. The first stage of the transform provides subband decomposition using wavelet transform instead of laplacian pyramid in the contourlet transform to remove the redundancy of the laplacian pyramid. And the second stage is a directional filter bank which produces angular decomposition at the same directions. Wavelet based contourlet transform brings a more sparse image expansion and can capture the intrinsic geometrical structure of images, which is more widely applied in image processing than wavelet transform and contourlet transform.The principle and implementation of wavelet based contourlet transform and its application on image fusion and image enhancement is systemically studied in this paper. The main work and achievements are as follows:1. The theory, implementation of 2-D separable wavelet transform was particularly studied. The disadvantage of 2-D separable wavelet transform in image processing and multi-scale geometrical analysis was particularly discussed. The theory, implementation of contourlet transform and wavelet based contourlet transform was particularly studied. The precedence of contourlet transform to wavelet transform and the precedence of wavelet based contourlet transform to contourlet transform were illustrated through non-linear approximation experiments.2.The theory of image denoising was particularly studied including its concepts and implementation algorithms. The design and the selection of the image denoising rule Was particularly researched and discussed. A method for image denoising based on wavelet based contourlet transform was proposed taking advantage of the transform's good property of image denoising. Combing the local energy matching rule, the method got better effect than wavelet transform and contourlet transform.3.The theory of image enhancement was particularly studied including its concepts and implementation algorithms. The design and the selection of the enhancing functions was particularly researched and discussed. A method for image enhancement based on wavelet based contourlet transform was proposed taking advantage of the transform's good property of multi-resolution and multi-direction. Combing enhancing function, the method got better effect than wavelet transform and contourlet transform to enhance the details of images.4.Shift-variance in wavelet based contourlet transform caused by critical down-sampling in the first stage wavelet transform results in visual artifacts called "Gibbs-like"phenomena. Cycle spinning was introduced to decrease the shift-variance and restrain the distortion and improved the image's quality.