基于Contourlet变换的图像降噪算法研究

发布时间：2018-03-25 22:50

本文选题：图像降噪　切入点：非下采样Contourlet变换(NSCT)　出处：《西南科技大学》2017年硕士论文

【摘要】：在传输和数字化过程中,图像往往受到成像设备和外界环境噪声的干扰,从而降低了数字图像的分辨率,偏离了图像本身的真实性。这不仅影响了人们的主观视觉效果,而且严重阻碍了后续图像的处理。因此,图像去噪是数字图像处理的基础和关键步骤。图像去噪技术的研究具有重要的理论意义和商业价值。本文的主要研究工作是利用非下采样Contourlet变换,在不同的噪声环境下实现更好的图像去噪效果。首先,以小波变换的基本理论作为研究的基础,分析了其在图像处理应用中的不足。针对不足,引入Contourlet变换的基本理论和构成的各种滤波器,分析了优点,也指出了不具有平移不变性而产生混频现象的缺陷,从而深入地探讨非下采样Contourlet变换的理论,来解决Contourlet变换中出现的伪Gibbs现象。其次,提出了在高斯白噪声的情况下,利用非下采样Contourlet变换,对于高频部分得到的系数,利用尺度间系数能量相关性构造的自适应阈值和改进的阈值函数共同对其进行处理,实现图像的降噪。进一步,根据实际情况,在高斯白噪声和椒盐噪声混合的情况下,提出对于低频部分的系数采用中值滤波降低椒盐噪声,以其提高图像的对比度。通过实验,利用MATLAB软件的仿真,对比了本论文提出方法的降噪效果和小波变换采用Bayes Shrink阈值的硬阈值函数和软阈值函数的降噪效果。通过主观和客观两方面的评价,得出本文所做的工作对图像的降噪效果有所提高。
[Abstract]:In the process of transmission and digitization, the image is often disturbed by the imaging equipment and the external environment noise, which reduces the resolution of the digital image and deviates from the authenticity of the image itself, which not only affects the subjective visual effect of people. And it seriously hinders the processing of subsequent images. Image denoising is the foundation and key step of digital image processing. The research of image denoising technology has important theoretical significance and commercial value. First of all, the basic theory of wavelet transform is taken as the basis of the research, and its shortcomings in the application of image processing are analyzed. This paper introduces the basic theory of Contourlet transform and various kinds of filters, analyzes the advantages, and points out the defects of mixing phenomenon without translation invariance. The theory of non-downsampling Contourlet transform is discussed in depth. In order to solve the pseudo-#en1# phenomenon in Contourlet transform. Secondly, it is proposed that under the condition of Gao Si white noise, using non-downsampling Contourlet transform, for the coefficient obtained in high frequency part, The adaptive threshold constructed by the energy correlation between scales and the improved threshold function are used to deal with the noise reduction of the image. Furthermore, according to the actual situation, under the condition of mixing Gao Si white noise with salt and pepper noise, In order to improve the contrast of the image, the median filter is used to reduce the salt and pepper noise for the low frequency part. Through the experiment, the simulation of MATLAB software is used. The noise reduction effect of the proposed method is compared with that of hard threshold function and soft threshold function using Bayes Shrink threshold in wavelet transform. It is concluded that the effect of image denoising is improved by the work done in this paper.
【学位授予单位】：西南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】