《基于小波包分析的F-X与K-L联合去噪研究和应用》由会员分享，可在线阅读，更多相关《基于小波包分析的F-X与K-L联合去噪研究和应用（84页珍藏版）》请在金锄头文库上搜索。
1、成都理工大学硕士学位论文基于小波包分析的F-X与K-L联合去噪研究和应用姓名：朱普申请学位级别：硕士专业：地球探测与信息技术指导教师：赵宪生20100501摘 要 I 基于小波包分析的基于小波包分析的 F-X与与 K-L联合去噪的研究联合去噪的研究和和应用应用 作者简介：朱普，男，1983 年 7 月生，师从成都理工大学赵宪生教授，2010年 6 月毕业于成都理工大学地球探测与信息技术专业，获得工学硕士学位。 摘摘 要要 伴随在有用信号中的噪声是影响地震信号处理的重要因素， 在不降低信号分辨率的基础上，降低信号中的噪声以及从强噪声中提取有用信号，就成了一个需要深入研究的课题。为了在不影响分辨率
2、的前提下提高信噪比，人们根据信号与噪声的各种特征差异，设计了许多去除噪声，提高信噪比的方法。 本论文以地震资料去噪方法为研究对象， 研究内容主要以围绕如何去除随机噪声进行。系统地对噪声进行了分析和阐述，针对地震资料去噪中所面临的各种噪声其特点、产生原因、影响程度和压制方法进行了研究。对目前实际生产中常用的一些去噪方法的原理、特点和适应性进行了系统的分析。 针对其中的 F-X 域去噪、K-L 变换和小波包去噪的进行具体的分析和研究，阐述了各自的去噪原理，并将各方法对地震信号进行模拟和实际资料处理。F-X域预测去噪技术是一项最基本的技术， 该技术旨在压制二维地震记录中的随机噪声，它以理论上的严密性
3、和实际效果上的显著性得到广泛应用。K-L 变换是作为一种特殊的正交线性变换，通过把地震道中的相干能量集中在有限几个主分量上，把相关性好的信号保存下来,从而滤除随机信号。实际处理结果表明,K-L 变换可以有效去除线性干扰,从而保留更多的浅层有效信息。小波分析是当前地震信号去噪中一个迅速发展的新领域， 而小波包的发展是频率域中小波的加细并且基于 Daubechies(分裂法)；因为小波包分析对上层的低频部分和高频部分同时进行分解，具有更加精确地局部分析能力。 根据 F-X 域去噪、K-L 去噪、和小波包分析去噪的各自特点，提出了联合去噪的方法。首先通过小波包分析对地震剖面进行分频处理，再根据小波系
4、数剖面在空间域的可以预测行进行 F-X 域去噪，选用合适的阀值对系数进行阀值去噪后，再将小波包系数进行重构，最后使用 K-L 变换提取地震信号中的相关信号。该方法不仅利用信号和噪声在小波包分析下奇异性截然不同的表现特征来去除噪声，还利用地震信号在空间上的可预测性和地震信号道间相关性进行去噪处理。通过将小波包分析法，F-X 域去噪和 K-L 变换进行有机结合使用，充分发挥各自的优势，达到较好的去噪效果。通过对合成理论模型实现和实际地震资料处理的反复实验验证表明，该方法与常规去噪方法相比，有更好的去噪处理效果，不仅能有效地去除强随机噪声，提高地震记录的信噪比和分辨率，而且对信号的保真度好，基本不会
5、使有效信号发生畸变。 关键词：小波包 去噪 F-X K-L 变换 Abstract III Research and Application of based on wavelet packet analysis KL and FX Joint denoising About the author: Zhu P, M, July 1983 Health, where he studied under Professor Zhao Xiansheng Chengdu University of Technology, June 2010 graduated from the Chengdu Un
6、iversity of Earth Observation and Information Technology, received a Master of Science. Abstract With the useful signal in noise is an important factor in seismic signal processing, without lowering the resolution on the basis of the signal and reduce signal noise and noise from the strong to extrac
7、t useful signals, it becomes a subject of further study. Does not affect the resolution in order to improve the SNR under the premise, people under a variety of signal and noise characteristics of different designs of many suppress noise, enhance signal to noise ratio method. This thesis Denoising o
8、f seismic data as the research object, research mainly focus on how to remove the random noise. Systematically analyzed and explained the noise for seismic data denoising faced by the noise of their characteristics, causes, impact, and methods of repression. The current actual production of some com
9、monly used denoising principle, characteristics and adaptability of the system. For which the FX domain noise suppression, KL transform and wavelet packet denoising for specific analysis and research, outlined their principles of noise reduction, and the method of the seismic signal simulation and r
10、eal data. FX forecast domain denoising technique is a fundamental technology that aims to suppress the two-dimensional random noise in seismic records, its close to theoretical and practical significance of results on widely used. KL transform as a special kind of orthogonal linear transformation, t
11、hrough the coherent seismic energy concentrated in a limited number of principal components, the correlation between a good signal to survive, thus filtered random signal. The results show that the actual processing, KL transform can effectively remove the linear interference, which retain more of t
12、he shallow effective information. Wavelet analysis is the current seismic signal denoising in a rapidly developing new areas of development of the wavelet packet frequency domain wavelet refinement and is based on Daubechies (split method); because the wavelet packet analysis of the upper part of th
13、e low frequency part and high frequency At the same time to break it down, with more accurate local analysis. According to FX denoisization, KL denoising, and wavelet packet analysis Denoising of their own characteristics to make a joint denoising methods. First, the wavelet packet analysis of the s
14、eismic profiles frequency processing, then the wavelet coefficients of profile in the space domain, can predict the row FX domain noise 成都理工大学硕士学位论文 IV suppression, choose the appropriate threshold on the coefficient of threshold de-noising, the wavelet packet coefficients then reconstructed, and fi
15、nally use the KL transform to extract the relevant signal in the seismic signals. The method makes use of signal and noise in wavelet packet analysis under different performance characteristics of Singularity to remove noise, also using seismic signals in space predictability and correlation between
16、 seismic signals Road denoising. By wavelet packet analysis, FX and the KL transform domain denoising using the combination, give full play to their strengths, achieve better denoising effect. Tong Guo theoretical models of synthetic and real seismic data processing to achieve the repeated experiments showed that the method and the conventional denoising methods, better denoising effect, not only can effectively remove strong random noise, imp
