Application Research Of Contourlet Transform In Image Processing

As separable wavelet transform has many demerits in image processing, contourlet transform technique is introduced in this thesis. Contourlet transform is composed of Laplacian Pyramid algorithm and directional filter banks. And, Laplacian Pyramid algorithm is one kind of multi-resolution analysis methods similar as separable wavelet transform, which can decompose the image into different resolution layers and achieve multi-resolution representation for images. Directional filter banks, with the capability of directional information selectivity, can capture directional information, which is one of the important characteristics of images. Besides, contourlet transform can also detect contour information, which is the intrinsic geometrical characteristic of images. Furthermore, with the characteristic of anisotropy, contourlet transform is very suitable for effective representation of images. Based on the analysis hereinbefore, we know that contourlet transform can capture not only the intrinsic geometrical structures but also the directional information of the image, so it can remedy the demerits of separable wavelet transform for image processing and can achieve effective and accurate representation of the image.In this thesis, we thoroughly studied the theory of contourlet transform, and investigated its application in image processing. These are embodied through the following aspects:(1) Based on the effective image representation of contourlet transform, the thresholding mechanism is introduced to achieve image de-noising. At the same time, the multi-layer thresholding mechanism is introduced in this thesis to achieve more effective image de-noising.(2) The SAR image de-speckling is investigated based on contourlet transform; And, the speckle-suppression algorithm, combining stationary wavelet transform, directional filter banks and the Bayesian Maximum a Posteriori estimation algorithm, is proposed and achieved great speckle reduction.(3) The image enhancement is accomplished by combining image representation of contourlet transform with simple but effective Bayesian nonlinear transform function.All the algorithms referred hereinbefore are implemented through experiments and compared with traditional separable wavelet transform to demonstrate the effectiveness and superiority of contourlet transform for image processing. The experimental results show that contourlet transform is satisfying for image de-noising and enhancement and is a potential technique for image processing.