基于多尺度几何分析的图像去噪研究-电子与通信工程专业论文.docxVIP

摘 摘 要 万方数据 万方数据 摘 要 随着通信技术和网络技术的发展以及各种数字电子产品的日益普及，图像 已成为人们获取外界信息的重要载体。然而图像在采集、传输和存储过程中很 容易受到噪声的干扰而导致图像质量降低，并且图像质量的好坏对后续处理有 着非常重要的影响。因此我们必须对图像进行降噪处理，以获得更高质量的图 像。 图像去噪方法主要分为空间域去噪和变换域去噪。变换域去噪是目前应用 非常广泛的方法，小波变换和多尺度几何变换是主要的变换方法。多尺度几何 变换相对小波变换具有很强的方向性，能取得更好的去噪效果，因此多尺度几 何变换去噪成为去噪领域研究的热点。 本文首先研究了多尺度几何分析的基本理论，重点研究了 Curvelet 变换的基 本理论及其实现方法。然后研究了 Curvelet 变换的硬阈值去噪和软阈值去噪，并 分别采用全局阈值和局部阈值进行实验，证明并分析了局部阈值能更好的适应 Curvelet 变换后图像能量分布的差异性。相对全局阈值去噪，局部阈值去噪能取 得更好的去噪效果。在局部阈值去噪的基础上，又分别从去噪阈值和阈值函数 的角度加以改进，实现了 Curvelet 自适应去噪和改进阈值函数的去噪方法。最后 将改进后的 Curvelet 去噪和全变差去噪联合起来实现去噪，即 Curvelet 变换和全 变差的联合图像去噪算法。相比之前的去噪算法，该方法取得了最好的去噪效 果，不但提高了去噪后的峰值信噪比，并改善了图像的视觉效果，因此是一种 有效的去噪算法。 关键词：图像去噪；小波变换；多尺度几何变换；Curvelet 变换；全变差 I AB ABSTRACT ABSTRACT With the development of communication technology and network technology, as well as the rising popularity of digital electronic products, the image has become an important carrier of access to outside information. However, the image in the process of collection, transmission and storage are vulnerable to the interference of noise which results in the decrease of image quality, and the image quality has very important influence for subsequent processing, therefore it is necessary for image noise reduction processing, in order to obtain higher quality images. Image denoising methods are divided into spatial domain denoising and transform domain denoising.Transform domain denoising is currently very popular denoising method, wavelet transform and multiscale geometric transformation is the major transformation methods. Relative to wavelet transform, multiscale geometric transformation is highly directional, can achieve better denoising effect, so multiscale geometric transformation denoising has become the research hotspot in the field of denoising. This paper first studied the basic theory of multiscale geometric analysis, mainly studied the basic theory of Curvelet transform and its realization method. Then studied hard threshold denoising and soft threshold denoising of Curv