Abstract: Multiwavelet is a new development in the wavelet theory and it can offer simultaneously orthogonality,symmetry,and short support. In signal processing,the orthogonality preserves energy,the symmetry avoids signal distortion and the short support reduces the boundary effects. Therefore multiwavelet is very suitable for various signal processing applications,especially denoising. The paper presents multiwavelet principles,transformation procedures,pre-processing methods and proposes a new GHM-like multiwavelet-based denoising method. In seismic data processing,the attenuation of random noise is an important research subject. Conventional temporal or spatial filtering methods often damage the useful signals while suppressing noise,and the single wavelet transforms can cause signal distortion since it fails to offer simultaneously the orthogonality and symmetry. The paper adopts multiwavelet and multiresolution method to remove noise contained in seismic data. Seismic data is first preprocessed to generate a group of vector data,and then approximate and detailed signals of various scales are generated by two-level multiwavelet transformation. Finally detailed signals are processed by soft threshold and denoised seismic data are obtained by reverse multiwavelet transformation. The denoising experiments of synthetic and real data show that multiwavelet transform is effective for noise reduction and can preserve signal features at the same time.

Key words: multiwavelet, pre-filter, de-noising, soft threshold

摘要： 多小波是对小波理论的一个新发展,它可以同时满足正交性、对称性、短支撑等良好的特性要求。文中介绍了多小波基本理论、多小波变换具体过程及预处理方法,提出了基于GHM类多小波变换的地震资料软阈值去噪方法,通过对合成数据和实际资料进行处理分析,表明多小波变换在有效压制随机噪声的同时,能较好地保留原信号的特征信息,是一种行之有效的去噪方法。

关键词: 多小波, 预处理, 去噪, 软阈值