基于双树轮廓波的图像降噪研究【毕业论文】.doc

- PAGE I - 本科毕业设计 （20 届） 基于双树轮廓波的图像降噪研究 所在学院 专业班级 电子信息科学与技术 学生姓名 学号 指导教师 职称 完成日期 年 月 PAGE I 摘　要 【摘要】轮廓波变换（Contourlet）具有良好的高维奇异性表达能力，在图像降噪中得到了广泛的应用，但其平移不变性的缺乏制约了性能的提高。本文借鉴双树复小波的先进思想，引入双树轮廓波变换（Dual Tree Contourlet，DT-Contourlet），通过将拉普拉斯塔形分解与双树方向滤波器组结合，使变换具有平移不变性，从而减轻乃至消除传统Contourlet降噪算法所带来的伪吉布斯现象。为了进一步提高降噪算法的性能，在分析VisuShrink阈值算法的基础上，采用双变量模型来刻画系数的尺度间相关性，并根据DT-Contourlet的复子带系数特性，在贝叶斯框架下实现了一种双变量阈值函数，提出一种基于DT-Contourlet变换的图像降噪新方法。实验结果表明，与传统方法相比，本文方法在降噪后图像的视觉效果及PSNR指标上都有较大提高。 【关键词】双树轮廓波；图像去噪；双变量模型；多尺度分析。  Abstract 【ABSTRACT】Contourlet has a good outline of the high singularity of expression, and has been widely used in image denoising , but its lack of translation invariance restricts the performance improvement. We compared to the dual tree complex wavelet transform than combine pyramid filters and direction filter banks to construct dual tree Contourlet transform. It is shiftable invariance and can eliminate pseudo-Gibbs phenomenon through the provision of complex sub-band coefficients when we conduct image denoising. In the specific methods we use VisuShrink estimated threshold, and appropriately improve it to achieve the best improvement. In order to meet dual tree profile, we improve the traditional, in the framework of the improved bivariate threshold function we propose bivariate model of dual tree Contourlet. Experimental results show that compared with wavelet denoising and contour denoising methods, the proposed method in the visual effects and objective evaluation (PSNR) both has a larger increase. 【KEYWORDS】Dual Tree Contourlet; Image Denoising; Bivariate Model; Multi-scale Analysis.  目　录 TOC \o 1-3 \h \z \u HYPERLINK \l _Toc293954041 摘　要 PAGEREF _Toc293954041 \h II HYPERLINK \l _Toc293954042 Abstract PAGEREF _Toc293954042 \h III HYPERLINK \l _Toc293954043 目　录 PAGEREF _Toc293954043 \h IV HYPERLINK \l _Toc293954044 1 绪论 PAGEREF _Toc293954