1 / 44

为了有效、快速地对图像进行处理和分析,常需要将定义在图像空间的图像以某种形式转换到其他空间中。

为了有效、快速地对图像进行处理和分析,常需要将定义在图像空间的图像以某种形式转换到其他空间中。 在这些空间中,可以利用它们特有的性质,方便地对转变后的图像进行处理,最后再转回图像空间以得到所需效果。. The purpose of image transformation : Simplify the problem of image processing, and improve the computing efficiency ( 简化图像处理问题,提高计算效率 ) Be favor to the features extracting of image

nuru
Download Presentation

为了有效、快速地对图像进行处理和分析,常需要将定义在图像空间的图像以某种形式转换到其他空间中。

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. 为了有效、快速地对图像进行处理和分析,常需要将定义在图像空间的图像以某种形式转换到其他空间中。为了有效、快速地对图像进行处理和分析,常需要将定义在图像空间的图像以某种形式转换到其他空间中。 在这些空间中,可以利用它们特有的性质,方便地对转变后的图像进行处理,最后再转回图像空间以得到所需效果。

  2. The purpose of image transformation: • Simplify the problem of image processing, and improve the computing efficiency (简化图像处理问题,提高计算效率) • Be favor to the features extracting of image 有利于图像特征提取。一些图像用频域信息进行分析,有助于从概念上增强对图像信息的理解。 • 更有效的存储和传输。

  3. (c) (a) Original image (b) Compared with the Fig(a), the skin fine lines can be reduced obviously in Fig(b) and (c) by using GLPF (Gaussian Lowpass Filter).

  4. (a) (b) (c) (d)

  5. 图像变换是许多图像处理和分析的基础。 Fourier Transform (FT)在图像处理和分析技术中,曾经起过并仍在起着重要的作用,被用于图像增强、复原、编码和描绘。 本章将着重介绍FT、FT反变换及其性质。

  6. Basic steps for filtering in the frequency domain

  7. Basic steps for filtering in the frequency domain • Multiply the input image by (-1)(M+N)/2 to center the transform to u=M/2 and v=N/2 傅立叶变换以零点为中心,导致谱图像最亮点在图像的左上角,为符合正常习惯,常将F(u,v)的零原点从图像的左上角移位到图像中心处。 为达到这一要求,根据傅立叶变换的性质,只需要将原图像乘以(-1)(M+N)/2,然后再进行傅立叶变化即可。 Note: if M and N are even numbers, then the shifted coordinates will be integers. • computer F(u,v), the DFT of the image from (1) 利用离散傅立叶变换,将图像从空间域中转换到频率域中

  8. Basic steps for filtering in the frequency domain • multiply F(u,v) by a filter function H(u,v) 对频率域中的图像进行特定处理 • compute the inverse DFT of the result in (3) 利用离散傅立叶发变换,将处理后的图像从频率域中反转换到空间域中 • obtain the real part of the result in (4) 反变换时,需要用到复数形式,为获取图像,必须将复数部分取消 • multiply the result in (5) by (-1)(M+N)/2 to cancel the multiplication of the input image. 重新定位图像原点

  9. 一、Fourier Transform • 1-D Continuous Fourier Transform • 2-D Continuous Fourier Transform • 1-D Discrete Fourier Transform • 2-D Discrete Fourier Transform • Some properties of Fourier Transform性质 二、几种常用的滤波函数H(u,v) • Smoothing frequency domain filters • Sharpening frequency domain filters

  10. 4.1.1 1-D Continuous Fourier Transform • 实变量x的连续可积函数f(x)的FT • 从F(u)中恢复f(x),定义为Inverse FT 记:

  11. 4.1.1 1-D Continuous Fourier Transform • FT—— • IFT—— • 幅度—— 幅度函数|F(u)|也被称为傅立叶谱 • 相角——

  12. 矩形函数的傅立叶谱 Example 1: 试求矩形函数f(x)的傅立叶变换F(u) 解:

  13. Example 2: 对高斯函数G(t),求其傅立叶变换。 解: 高斯函数的傅立叶变换同样是高斯函数

  14. 4.1.1 1-D Continuous Fourier Transform 常见函数的1-D CFT:

  15. 4.1.2 2-D Continuous Fourier Transform • 实变量x,y的连续可积函数f(x,y)的傅立叶变换 • 从F(u,v)中恢复f(x,y),定义为傅立叶反变换 记:

  16. 4.1.2 2-D Continuous Fourier Transform • FT—— • IFT—— • 幅度—— • 能量谱—— • 相角——

  17. Example 3: 求二维矩形函数的傅立叶变换。 解:

  18. 4.1.3 1-D Discrete Fourier Transform (DFT) 对连续可积函数f(x),取间隔x,将f(x)离散化为一个序列:{f(x0),f(x0+x), …, f(x0+(N-1)x)} 将序列表示为:f(x)=f(x0+xx)

  19. 4.1.3 1-D DFT • 被抽样函数的DFT可表示为 • DFT反变换

  20. 设:N=4

  21. DFT的矩阵表示法 记作:F=Wf

  22. 4.1.4 2-D DFT 将1-D DFT推广到二维 在数字图像处理中,图像一般取方形,即:M=N

  23. 4.1.5 傅立叶变换的主要性质 • 加法定理 • 位移定理 • 卷积定理 • 可分离性

  24. 4.1.5 FT的主要性质——加法定理 时域或空域内的相加对应于频域内的相加。 设有两个傅立叶对: 若:r(t)=f(t)+g(t) 则:

  25. 4.1.5 FT的主要性质——位移定理 函数位移不改变Fourier Transform的幅值。

  26. 位移定理 4.1.5 FT的主要性质——卷积定理 时域(或空域)中的卷积等价于频域的乘积。

  27. f(x,y) F(x,v) F(u,v) 4.1.5 FT的主要性质——可分离性 1个2-D DFT等价于2个1-D DFT。 1-D Row transform 1-D Column transform

  28. 4.2 几种常用的滤波函数H(u,v) • Ideal Lowpass Filter(ILPF) • Butterworth Lowpass Filter(BLPF) • Gaussian Lowpass Filter(GLPF) • Ideal Highpass Filter • Butterworth Highpass Filter(BHPF) • Gaussian Highpass Filter(GHPF) Smoothing Sharpening

  29. 4.2.1 ILPF • Perspective (透视的) plot of an ILPF • Filter diaplayed as an image • Filter radial cross section

  30. 4.2.1 ILPF D0——截止频率 D(u,v)=(u2+v2)1/2——频率平面原点到点(u,v)的距离

  31. 4.2.1 ILPF 特点: • 物理上不可实现 • 有抖动现象 • 滤除高频成分,使图像变模糊

  32. Lowpass frequency domain filtering; for the original image and its spectrum: • Spectrum of a low-pass filtered image, all higher frequencies are filtered out • image resulting from the inverse Fourier transform applied to spectrum a • Spectrum of a low-pass filtered image, only very high frequencies are filtered out, • inverse Fourier transform applied to spectrum c. Example of ILPF

  33. 4.2.2 Butterworth Lowpass Filter (BLPF) • Perspective (透视的) plot of an BLPF • Filter diaplayed as an image • Filter radial cross section of orders 1 through 4

  34. 4.2.2 BLPF • n为阶数 • 当D(u,v)=D0时,H(u,v)降为最大值的一半。

  35. 剖面图 1阶BLPF转移函数 三维图

  36. 3阶BLPF转移函数 三维图 剖面图

  37. 4.2.3 Gaussian Lowpass Filter (GLPF) • Perspective (透视的) plot of an GLPF • Filter diaplayed as an image • Filter radial cross section for various values of D0

  38. Example of GLPF • Sample text of poor resolution (note broken characters in magnified 放大的 view) • Result of filtering with a GLPH (broken character segments were joined)

  39. 4.2.4 Sharpening Frequency Domain Filter

  40. IHPF转移函数 三维图 剖面图

  41. 3阶BHPF转移函数 三维图 剖面图

  42. Example of IHPF Results of IHPF with D0=15, 30 and 80 respectively. Problems with ringing are quite evident in (a) and (b).

  43. Example of BHPF Results using BHPF of order 2 with D0=15, 30 and 80 respectively. These results are much smoother than those obtained with an ILPF.

  44. BLPF (a) is the original image. (b)-(f) the results using BHPF of order 2 with cutoff frequencies at radii of 5, 15, 30, 80 and 230 respectively.

More Related