Message Board

Dear readers, authors and reviewers,you can add a message on this page. We will reply to you as soon as possible!

2022 Volume 47 Issue 12
Article Contents

ZHAO Yanbo, FENG Qiang. Linear Canonical Wavelet Convolutions and Their Properties[J]. Journal of Southwest China Normal University(Natural Science Edition), 2022, 47(12): 70-77. doi: 10.13718/j.cnki.xsxb.2022.12.007
Citation: ZHAO Yanbo, FENG Qiang. Linear Canonical Wavelet Convolutions and Their Properties[J]. Journal of Southwest China Normal University(Natural Science Edition), 2022, 47(12): 70-77. doi: 10.13718/j.cnki.xsxb.2022.12.007

Linear Canonical Wavelet Convolutions and Their Properties

More Information
  • Corresponding author: FENG Qiang
  • Received Date: 22/06/2022
    Available Online: 20/12/2022
  • MSC: O241.86;TN911.7

  • Linear canonical wavelet transform is the characteristic of combining linear canonical transform and wavelet transform characterization, which can realize the multi-resolution analysis of signal in the time-linear canonical domain, processing more complex signals. It has been proved to be a powerful tool in signal processing field. People grasp the characteristics of linear canonical transformation relatively late. Linear canonical transformation expands the limitations of the wavelet transform, linear canonical transformation has the three additional free parameters, it is more flexible and often used in time-frequency analysis and non-stationary signal processing, retained the wavelet transform multiple resolution analysis. However, its theoretical system is still incomplete, and there are many theories related to signal processing that need to be improved. We first give the tolerance and canonical conditions for linear canonical wavelet functions, and then study a new class of linear canonical wavelet convolution and correlation theorems. Finally, we use the obtained theorem to study the filtering design of linear canonical wavelet domain.
  • 加载中
  • [1] MOSHINSKY M, QUESNE C. Linear Canonical Transformations and Their Unitary Representations[J]. Journal of Mathematical Physics, 1971, 12(8): 1772-1780. doi: 10.1063/1.1665805

    CrossRef Google Scholar

    [2] COLLINS S A. Lens-System Diffraction Integral Written in Terms of Matrix Optics[J]. Journal of the Optical Society of America, 1970, 60(9): 1168-1177. doi: 10.1364/JOSA.60.001168

    CrossRef Google Scholar

    [3] 许天周, 李炳照. 线性正则变换及其应用[M]. 北京: 科学出版社, 2013.

    Google Scholar

    [4] COOLEY J W, TUKEY J W. An Algorithm for the Machine Calculation of Complex Fourier Series[J]. Mathematics of Computation, 1965, 19(90): 297-301. doi: 10.1090/S0025-5718-1965-0178586-1

    CrossRef Google Scholar

    [5] OZAKTAS H M, KUTAY M A. The Fractional Fourier Transform[C]//2001 European Control Conference (ECC). September 4-7, 2001. Porto. IEEE, 2001: 1477-1483.

    Google Scholar

    [6] JAMES D F V, AGARWAL G S. The Generalized Fresnel Transform and Its Application to Optics[J]. Optics Communications, 1996, 126(4-6): 207-212. doi: 10.1016/0030-4018(95)00708-3

    CrossRef Google Scholar

    [7] DENG B, TAO R, WANG Y. Convolution Theorems for the Linear Canonical Transform and Their Applications[J]. Science in China Series F: Information Sciences, 2006, 49(5): 592-603. doi: 10.1007/s11432-006-2016-4

    CrossRef Google Scholar

    [8] WEI D Y, RAN Q W, LI Y M, et al. A Convolution and Product Theorem for the Linear Canonical Transform[J]. IEEE Signal Processing Letters, 2009, 16(10): 853-856. doi: 10.1109/LSP.2009.2026107

    CrossRef Google Scholar

    [9] 王荣波, 冯强. 线性正则正弦与余弦变换的卷积定理及其应用[J]. 光电工程, 2018, 45(6): 73-82.

    Google Scholar

    [10] ZHANG Y N. Uncertainty Principle for the Quaternion Linear Canonical Transform in Terms of Covariance[J]. Journal of Beijing Institute of Technology, 2021, 30(3): 238-243.

    Google Scholar

    [11] 宋浒, 张利, 许梦晗, 等. 基于形态学和小波变换的图像边缘检测方法[J]. 西南大学学报(自然科学版), 2020, 42(4): 105-111.

    Google Scholar

    [12] LI Y X, JIAO S B, GAO X. A Novel Signal Feature Extraction Technology Based on Empirical Wavelet Transform and Reverse Dispersion Entropy[J]. Defence Technology, 2021, 17(5): 1625-1635. doi: 10.1016/j.dt.2020.09.001

    CrossRef Google Scholar

    [13] 苏博妮. 基于梯度域的光场全聚焦图像生成方法[J]. 西南大学学报(自然科学版), 2020, 42(10): 174-182.

    Google Scholar

    [14] PÉREZ-RENDÓN A F, ROBLES R. The Convolution Theorem for the Continuous Wavelet Transform[J]. Signal Processing, 2004, 84(1): 55-67. doi: 10.1016/j.sigpro.2003.07.014

    CrossRef Google Scholar

    [15] SHI J, ZHANG N T, LIU X P. A Novel Fractional Wavelet Transform and Its Applications[J]. Science China Information Sciences, 2012, 55(6): 1270-1279. doi: 10.1007/s11432-011-4320-x

    CrossRef Google Scholar

    [16] WEI D Y, LI Y M. Generalized Wavelet Transform Based on the Convolution Operator in the Linear Canonical Transform Domain[J]. Optik, 2014, 125(16): 4491-4496. doi: 10.1016/j.ijleo.2014.02.021

    CrossRef Google Scholar

    [17] 鲁媛媛. 分数阶小波变换相关的卷积定理[D]. 北京: 北京理工大学, 2015.

    Google Scholar

    [18] 杨秀竹, 余波. 3阶B-样条小波的希尔伯特变换的消失矩[J]. 西南师范大学学报(自然科学版), 2021, 46(12): 31-36.

    Google Scholar

    [19] WEI D Y, RAN Q W, LI Y M. A Convolution and Correlation Theorem for the Linear Canonical Transform and Its Application[J]. Circuits, Systems, and Signal Processing, 2012, 31(1): 301-312.

    Google Scholar

  • 加载中
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Figures(1)

Article Metrics

Article views(547) PDF downloads(301) Cited by(0)

Access History

Other Articles By Authors

Linear Canonical Wavelet Convolutions and Their Properties

    Corresponding author: FENG Qiang

Abstract: Linear canonical wavelet transform is the characteristic of combining linear canonical transform and wavelet transform characterization, which can realize the multi-resolution analysis of signal in the time-linear canonical domain, processing more complex signals. It has been proved to be a powerful tool in signal processing field. People grasp the characteristics of linear canonical transformation relatively late. Linear canonical transformation expands the limitations of the wavelet transform, linear canonical transformation has the three additional free parameters, it is more flexible and often used in time-frequency analysis and non-stationary signal processing, retained the wavelet transform multiple resolution analysis. However, its theoretical system is still incomplete, and there are many theories related to signal processing that need to be improved. We first give the tolerance and canonical conditions for linear canonical wavelet functions, and then study a new class of linear canonical wavelet convolution and correlation theorems. Finally, we use the obtained theorem to study the filtering design of linear canonical wavelet domain.

  • 线性正则变换[1-3]是一种比较新颖的、功能强大的信号处理工具,因其有3个自由参数,使得线性正则变换具有更多的灵活性. 傅里叶变换[4]、分数阶傅里叶变换[5]、菲涅耳变换[6]都是线性正则变换的特殊情形. 近年来,线性正则变换在解决光学系统、滤波器设计、时频分析等方面都有重要的应用[7-10].

    小波变换在信号处理领域是重要的时频分析工具,在时频域分析、表征图像边缘检测、图像处理、全聚焦图像生成等方面有广泛的应用[11-13]. 近年来,小波变换在信号处理中的理论与应用也逐渐引起国内外学者的重视.文献[14]推导出了小波变换的卷积和相关定理,表明其在时频上具有相似性,并且证明了卷积定理对加噪信号恢复的有效性. 文献[15]利用一种新型卷积,给出了均匀采样和低通重构公式. 文献[16]提出了广义小波变换的卷积定理. 文献[17]提出了分数阶小波变换的卷积和相关定理. 文献[18]研究了小波的希尔伯特变换,得到了更高的消失矩. 由于线性正则小波变换是分数阶傅里叶小波变换的进一步拓展,可以实现对信号在时间线性正则域的多分辨率分析,因此在信号处理中有着非常重要的作用.

    本文在线性正则变换与小波变换的基础上,首先给出了线性正则小波函数的容许性条件与正则性条件,其次研究了一类新型线性正则小波卷积与相关定理,最后利用所得定理,研究了线性正则小波域的滤波设计.

1.   预备知识
  • 定义1[3]  函数f(x)的线性正则变换定义为

    其中,核函数为

    M=(ABCD)为参数矩阵,ABCD$ \mathbb{R}$,且满足ADBC=1.

    线性正则变换的逆变换表示为

    其中,参数矩阵M-1=(D,-B,-CA).

    当参数矩阵M=(0,1,-1,0)时,线性正则变换退化为经典的傅里叶变换

    性质1(叠加性)  $L^{\left(A_2, B_2, C_2, D_2\right)}\left(L^{\left(A_1, B_1, C_1, D_1\right)}(f(x))\right)=L^{(E, F, G, H)}(f(x)) $.

    性质2(可逆性)  $L^{(D, -B, -C, A)}\left(L^{(A, B, C, D)}(f(x))\right)=f(x) $.

    定义2  设$f \in L^2 { (\mathbb{R}) } $,小波函数$\psi \in L^2({ \mathbb{R} }) $,且满足可容许性条件

    f的小波变换定义为

    其中,小波基函数定义为

    $a \in \mathbb{R}_{+}, b \in \mathbb{R} $分别代表尺度和平移参数. 小波变换的逆变换为

    定义3[16]  线性正则小波变换定义为

    这里的ψMab(x)为核函数,满足

    它的逆变换定义为

    M=(cos θ,sin θ,-sin θ,cos θ)时,线性正则小波变换退化为分数阶小波变换

    定义4[19]  线性正则变换的卷积运算定义为

    当参数矩阵M=(0,1,-1,0)时,线性正则卷积运算退化为经典的卷积运算

    基于传统的卷积算子,新的卷积运算可以由经典卷积运算表示为

    定义5[14]  设fg是定义在$\mathbb{R}^2 $上的复值函数,则仅有一个变量的卷积运算分别定义为

    定义6[19]  线性正则变换的相关运算定义为

    当参数矩阵M=(0,1,-1,0)时,线性正则变换的相关运算退化为

    定义7[14]  如果fg是定义在$\mathbb{R}^2 $上的复值函数,仅有一个变量的相关形式分别为

    定义8[14]  给定函数$f \in L^2(\mathbb{R}), h \in L^2\left(\mathbb{R}^2\right) $,它们的广义卷积定义为

    引理1[14]  设$\psi_f \in L^2(\mathbb{R}) \text { 和 } \psi_h \in L^1(\mathbb{R}) \cap L^2(\mathbb{R}) $是两个容许性小波,令$W_f^{\psi_f}, W_h^{\psi_h} $分别表示函数$f \in L^2(\mathbb{R}), h \in L^1(\mathbb{R}) \cap L^2(\mathbb{R}) $具有母小波ψfψh的小波变换,如果$g=(f * h), \psi_g=\left(\psi_f * \psi_h\right) $,那么就有

    引理2[17]  设$\psi_f \in L^2(\mathbb{R}) \text { 和 } \psi_h \in L^1(\mathbb{R}) \cap L^2 { (\mathbb{R}) } $是两个容许性小波,令$W_{f, \alpha}^{\psi_f}, W_{h, \alpha}^{\psi_h} $分别表示函数$f \in {L^2}(\mathbb{R}), h \in {L^1}(\mathbb{R}) \cap {L^2}{\rm{ (}}\mathbb{R}{\rm{)}} $具有母小波ψfψh的分数阶小波变换,如果$g=(f * h), \psi_g=\left(\psi_f * \psi_h\right) $,那么就有

    引理3[14]  设$\psi_f \in L^2(\mathbb{R}), \psi_h \in L^1(\mathbb{R}) \cap L^2(\mathbb{R}) $是两个容许性小波,令$W_f^{\psi_f}, W_h^{\psi_h} $分别表示函数$f \in L^2 { (\mathbb{R}) }, h \in L^1 { (\mathbb{R}) } \cap L^2 { (\mathbb{R}) } $具有母小波ψfψh的小波变换,如果$g=(f \odot h), \psi_g=\left(\psi_f \odot \psi_h\right) $,那么就有

    引理4[17]  设$\psi_f \in L^2(\mathbb{R}), \psi_h \in L^1(\mathbb{R}) \cap L^2(\mathbb{R}) $是两个容许性小波,令$W_{f, \alpha}^{\psi_f}, W_{h, \alpha}^{\psi_h} $分别表示函数$f \in L^2 { (\mathbb{R}) }, h \in L^1 { (\mathbb{R}) } \cap L^2 { (\mathbb{R}) } $具有母小波ψfψh的分数阶小波变换,如果$g=(f \odot h), \psi_g=\left(\psi_f \odot \psi_h\right) $那么就有

    定义9[14]  设$\psi \in L^2(\mathbb{R}) $为容许性小波,且有0≤ρ < N,若

    则称Mψρ为小波函数ψN阶消失矩.

    引理5[14]  设$\psi_f \in L^2(\mathbb{R}), \psi_h \in L^1(\mathbb{R}) \cap L^2(\mathbb{R}) $是两个具有N1N2阶消失矩的容许性小波,令$\psi_p=\left(\psi_f * \psi_h\right) \text { 和 } \psi_q=\left(\psi_f \odot \psi_h\right) $,则ψpψq也是可容许性小波并且具有Nf+Nh阶消失矩.

2.   主要研究内容
  • 定理1  设$\psi_f \in L^2(\mathbb{R}) \text { 和 } \psi_h \in L^1 { (\mathbb{R}) } \cap L^2 { (\mathbb{R}) } $是两个具有N1N2阶消失矩的容许性小波,设${\psi _p} = \left( {{\psi _f}\mathop \otimes \limits^A {\psi _h}} \right), {\psi _q} = \left( {{\psi _f}\mathop \odot \limits^A {\psi _h}} \right) $,则ψpψq也是可容许性小波并且具有Nf+Nh阶消失矩.

      由于$\psi_f \in L^2(\mathbb{R}), \psi_h \in L^1(\mathbb{R}) \cap L^2(\mathbb{R}) $,可知$\psi_p, \psi_q \in L^2(\mathbb{R}) $,且有

    因为ψfψh是容许性小波,根据(1)式知道0 < ψfψh < ∞,根据线性正则变换的性质,得到

    则可以得到

    因此,我们得出

    因为$\psi_h \in L^1(\mathbb{R}) $,则$L_{{\psi _h}}^\mathit{\boldsymbol{M}}(w) $有界,因此存在正实数K,使得$\left| {L_{{\psi _h}}^\mathit{\boldsymbol{M}}(w)} \right| \le K $,由线性正则变换与傅里叶变换的关系可得

    其中$\psi_h=\exp \left(\frac{j A x^2}{2 B}\right) \psi_f $. 因此小波ψpψq满足容许性条件.

    下面考虑ψpρ阶矩

    由定义4可得

    其中

    ρNf+Nh,若kNf,则有$ \boldsymbol{M}_{\psi_f}^k=0$,进而可得$\boldsymbol{M}_{\psi_f^*}^k=0 $,即有$ \boldsymbol{M}_{\psi_p}^\rho=0$. 若ρkNh,则有$\boldsymbol{M}_{\psi_h}^{\rho-k}=0 $,进而可得$\boldsymbol{M}_{\psi_h^*}^{\rho-k}=0 $,即$ \boldsymbol{M}_{\psi_p}^\rho=0$. 因此ψp具有Nf+Nh阶消失矩. 同理可证ψq具有Nf+Nh阶消失矩,证毕.

    定理2  设$\psi_f \in L^2 { (\mathbb{R}) 和 } \psi_h \in L^1 { (\mathbb{R}) } \cap L^2 { (\mathbb{R}) } $是两个容许性小波,令$W_{f, \boldsymbol{M}}^{\psi_f}, W_{h, \boldsymbol{M}}^{\psi_h} $分别表示函数$f \in L^2(\mathbb{R}), h \in L^1(\mathbb{R}) \cap L^2(\mathbb{R}) $具有母小波ψfψh的线性正则小波变换,如果

    则有

    g(x)的线性正则小波可以写为

    其中

    $g(x), \psi_g\left(\frac{x-b}{a}\right) $带入$W_{g, \boldsymbol{M}}^{\psi_g}(a, b) $中,得到

    $\alpha=x-\tau, \beta=b+a y-\tau $,且满足$\beta^2-b \beta=\frac{y(x-b)}{a}-\tau y $,即

    定理3  设$\psi_f \in L^2(\mathbb{R}) \text { 和 } \psi_h \in L^1 { (\mathbb{R}) } \cap L^2 { (\mathbb{R}) } $是两个容许性小波,令$W_{f, \boldsymbol{M}}^{\psi_f}, W_{h, \boldsymbol{M}}^{\psi_h} $分别表示函数$ f \in {L^2}(\mathbb{R}), h \in {L^1}{(\mathbb{R}) } \cap {L^2}{ (\mathbb{R})}$具有母小波ψfψh的线性正则小波变换,如果

    则有

      g(x)的线性正则小波可以写为

    其中

    $g(x), \psi_g\left(\frac{x-b}{a}\right) $带入$ W_{g, \boldsymbol{M}}^{\psi_g}(a, b)$中,得到

    $\alpha=\tau-x, \beta=\tau-b-a y, \gamma=\tau x-b^2+2 b $,且满足

    注1  定理1与定理3表明,线性正则小波卷积运算与相关运算在每个尺度上是独立的,结合线性正则小波变换可以在时频域联合表征信号特性,我们可以对给定的信号的不同尺度构造相应的空变滤波.

    注2  这些定理表明,两个信号卷积的联合时频表征可以表示为两个信号在每个固定频率上的联合时频表征在时间变量或空间变量上的卷积.

    由于联合时频表征的空变滤波不同于傅里叶域的时不变滤波,因此我们可以通过以下步骤实现时变或者空变滤波:

    步骤1  给信号f(x)做线性正则小波变换$W_\boldsymbol{M}^\psi(a, b) $

    步骤2  $W_\boldsymbol{M}^\psi(a, b) $乘传递函数T(ab);

    步骤3  给$W_\boldsymbol{M}^\psi(a, b) T(a, b) $做逆线性正则小波变换.

    空变滤波的实现如图 1所示:

    下面我们给出具体分析:

    $ f(x) \in L^2(\mathbb{R}), \psi_1, \psi_2 \in L^1(\mathbb{R}) \cap L^2(\mathbb{R}), T(a, b) \in L^2\left(\mathbb{R}^2\right), W_\boldsymbol{M}^{\psi_1}(a, b)$表示信号f(x)的具有小波函数ψ1的线性正则小波变换,

    对(2)式做逆线性正则小波变换,可得

    则可得线性正则小波变换域空变滤波

3.   结论
  • 本文在线性正则变换与小波变换卷积的基础上,研究了一类新型线性正则小波变换的卷积定理与相关定理. 首先给出了线性正则小波卷积和相关的容许性条件与正则条件;其次推导出线性正则小波变换的卷积定理;最后,利用所得卷积及其卷积定理,研究了线性正则小波变换域的滤波设计,给出了线性正则小波域的空变滤波的设计方法.

Figure (1)  Reference (19)

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return