二维离散傅里叶变换的平移性质证明

0 前言

本文记录了笔者数字图像处理作业的一道二维离散傅里叶变换的平移性质的证明题，希望能够帮助到读者。

1 证明

$$\begin{array}{rl}f(x,y)e^{j2\pi (\frac{u_{0}x}{M}+\frac{v_{0}y}{N})}& <=>F(u-u_0,v-v_0)\\ f(x-x_0,y-y_0)& <=>F(u,v)e^{-j2\pi (\frac{u_{0}x}{M}+\frac{u_{0}y}{N})}\end{array}$$
也就是说，用指数项乘以 $f(x,y)$ 将使DFT的原点移到 $(u_0,v_0)$；反之，用负指数乘以 $F(u,v)$ 将使 $f(x,y)$ 的原点平移到点$(x_0,y_0)$。即图像平移不影响幅度谱，但影响相位谱。

1.1 空间位移

$$f(x-x_0,y-y_0)<=>F(u^{\prime },v^{\prime })$$

$$F(u^{\prime },v^{\prime })=\frac{1}{\sqrt{MN}}\sum _{x=0}^{M-1}\sum _{y=0}^{N-1}f(x-x_0,y-y_0)e^{-j2\pi (\frac{\mathrm{u}x}{M}+\frac{\mathrm{v}y}{N})}$$

令 $x-x_0=a,y-y_0=b$
则
$$\begin{array}{rl}F(u^{\prime },v^{\prime })& \begin{array}{r}=\frac{1}{\sqrt{MN}}\sum _{x=0}^{M-1}\sum _{y=0}^{N-1}f(a,b)e^{-j2\pi (\frac{u(a+x_0)}{M}+\frac{\nu (b+y_0)}{N})}\end{array}\\ & =e^{-j2\pi (\frac{u{x}_0}{M}+\frac{v{y}_0}{N})}\cdot \frac{1}{\sqrt{MN}}\sum _{x=0}^{M-1}\sum _{y=0}^{N-1}f(a,b)e^{-j2\pi (\frac{ua}{M}+\frac{vb}{N})}\\ & =e^{-j2\pi (\frac{u{x}_0}{M}+\frac{v{y}_0}{N})}F(u,v)\end{array}$$
即
$$f(x-x_0,y-y_0)<=>F(u,v)\cdot e^{-j2\pi (\frac{u_{0}x}{M}+\frac{u_{0}y}{N})}$$

1.2 频率位移

$$F(u-u_0,v-v_0)<=>f(x^{\prime },y^{\prime })$$
$$f(x^{\prime },y^{\prime })=\frac{1}{\sqrt{MN}}\sum _{x=0}^{M-1}\sum _{y=0}^{N-1}F(u-u_0,v-v_0)e^{j2\pi (\frac{\mathrm{u}x}{M}+\frac{\mathrm{v}y}{N})}$$
令 $u-u_0=a,v-v_0=b$
则
$$\begin{array}{rl}f(x^{\prime },y^{\prime })& =\frac{1}{\sqrt{MN}}\sum _{x=0}^{M-1}\sum _{y=0}^{N-1}f(a,b)e^{j2\pi (\frac{(a+u_0)x}{M}+\frac{(b+v_0)y}{N})}\\ & =e^{j2\pi (\frac{u{x}_0}{M}+\frac{\nu y_0}{N})}\cdot \frac{1}{\sqrt{MN}}\sum _{x=0}^{M-1}\sum _{y=0}^{N-1}f(a,b)e^{j2\pi (\frac{ax}{M}+\frac{by}{N})}\\ & =e^{j2\pi (\frac{u{x}_0}{M}+\frac{v{y}_0}{N})}f(x,y)\end{array}$$
即
$$F(u-u_0,v-v_0)<=>f(x,y)\cdot e^{j2\pi (\frac{u_{0}x}{M}+\frac{u_{0}y}{N})}$$
综上，得证！

2 总结

以上就是笔者要分享的内容了，希望可以帮助到您，倘若大佬发现问题，直接评论区斧正即可，我们下期再见！🌞🌞🌞