基于中值滤波和小波变换的图像去噪算法研究

发布时间：2018-01-14 08:42

本文关键词：基于中值滤波和小波变换的图像去噪算法研究　出处：《河北大学》2014年硕士论文　论文类型：学位论文

【摘要】：在数字化信息时代的今天，图像信息已经成为一种极其重要的获取和传递信息的媒体和方式，而图像信号在产生、运输和记录的过程中，由于成像系统、传输介质和记录设备等不完善，使其受到噪声干扰而变得模糊，为后续图像的处理（如图像分割和特征提取等）带来很大不便。因此，图像的预处理（去噪和增强）则显得尤为重要。本文主要针对图像中最常见的脉冲噪声和高斯噪声，利用中值滤波和小波变换技术滤除进行了深入的探讨和研究，研究工作主要包括以下几个方面：首先，本文研究了标准的中值滤波、改进的中值滤波以及极值中值滤波算法，在比较分析它们各自的优缺点之后，提出了一种改进的极值中值滤波算法，并通过仿真比较试验，结果显示这种算法具有良好的图像细节保护性能，有较好的去噪效果。其次，本文研究了小波图像去噪方法，通过分析自然图像的小波系数的特点，针对小波阈值滤波技术中的阈值函数和阈值的选取进行深入研究之后，对受高斯噪声污染的像素点采用改进的阈值函数和阈值，去除高斯噪声。仿真结果表明，该算法去除高斯白噪声效果优于其他一些小波去噪算法。最后，本文针对某一种去噪算法通常只能对某一类噪声的去除有效，为了抑制混合噪声，，本文提出了基于极值中值滤波和基于高斯混合模型的小波去噪算法相结合的图像去噪方法，通过仿真实验证明，该算法的去噪效果比单独使用中值滤波和小波去噪具有更好的实用性和有效性。
[Abstract]:In the digital information age of today, the image information has become an extremely important way of media and the acquisition and transmission of information, and the image signal in the process of production, transportation and record, because the imaging system, transmission media and recording equipment is not perfect, which influenced by the noise and blur, for subsequent image processing the (such as image segmentation and feature extraction) inconvenience. Therefore, image preprocessing (denoising and enhancement) is particularly important.
In this paper, the most common impulse noise and Gauss noise in image are filtered and removed by median filter and wavelet transform.
Firstly, this paper studies a standard median filter, improved median filter and extremum median filtering algorithm, after comparing their advantages and disadvantages, and proposes an improved algorithm of extremum median filtering, and compared by simulation test, results show that the algorithm has good image detail protection performance, has a better denoising effect.
Secondly, this paper studies the wavelet image denoising method, through the analysis of characteristics of wavelet coefficients of natural images, the wavelet threshold filtering technique of threshold function and threshold of in-depth study, the pixels corrupted by Gauss noise by threshold function and threshold improvement, the removal of Gauss noise. The simulation results show that the this algorithm is better than the Gauss white noise removal effect of some other wavelet denoising algorithm.
Finally, in this paper a denoising algorithm of a class usually can only remove noise effectively, in order to suppress mixed noise, this paper presents an image median filter and wavelet based on Gauss mixture model denoising algorithm combining denoising method based on the simulation experiments, the validity and practicability of the denoising effect the algorithm is better than using median filter and wavelet denoising separately.

【学位授予单位】：河北大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN911.4

【参考文献】