小波理论与粒子群相结合的图像去噪研究

发布时间：2018-05-04 23:37

本文选题：图像滤波 + 小波理论　；参考：《新疆大学》2017年硕士论文

【摘要】：数字信号滤波一直是信息处理研究的一个热点,特别是近50年内各种新理论、新方法的提出,使信息处理成为二十世纪发展最快的学科之一。小波分析理论是现阶段处理信号最有效的具有时频特性的处理工具。本文重点研究了小波域的非线性阈值函数去噪。基于小波分析的图像去噪方法是图像去噪中最常用的方法之一。通过小波变换对图像进行多层分解,再由有效信息和干扰信息在小波域的奇异性可知,有效信息呈中低频,而干扰信息呈高频。根据这一特性,常在干扰信息处设定阈值对小波系数进行判断,将大于阈值的信息保留,而小于阈值的清零或做其他方式的处理。构造新的多层阈值函数,并选取固定估计阈值为阈值判定准则并对其做了优化改进,改善了固定估计阈值选取过大的问题。通过粒子群优化算法与新多层阈值函数相结合的方法,实现一种自适应的图像去噪算法。通过选取不同国际标准测试图像、不同类型和不同方差的噪声以及两种及其以上的混合噪声来进行实验仿真,通过与其他学者提出的阈值函数进行主观和客观的数据对比,验证本文方法在图像去噪上的有效性。
[Abstract]:Digital signal filtering has been a hot topic in information processing, especially in the past 50 years, with the development of new theories and methods, information processing has become one of the fastest developing subjects in the 20th century. Wavelet analysis theory is the most effective signal processing tool with time-frequency characteristics. In this paper, the nonlinear threshold function denoising in wavelet domain is studied. Image denoising based on wavelet analysis is one of the most commonly used methods in image denoising. The wavelet transform is used to decompose the image in multilayer, and from the singularity of the effective information and the interference information in the wavelet domain, it can be known that the effective information is middle and low frequency, while the interference information is high frequency. According to this characteristic, the wavelet coefficients are usually judged by setting a threshold at the interference information. The information greater than the threshold is preserved, while the information less than the threshold is cleared or processed in other ways. A new multilayer threshold function is constructed, and the fixed estimation threshold is selected as the threshold decision criterion, and the optimization improvement is made to improve the problem that the fixed estimation threshold is too large. An adaptive image denoising algorithm is implemented by combining particle swarm optimization with a new multilayer threshold function. By selecting different international standard test images, different types and variances of noise, and two or more kinds of mixed noise to carry out experimental simulation, the subjective and objective data are compared with the threshold function proposed by other scholars. The effectiveness of this method in image denoising is verified.
【学位授予单位】：新疆大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】