【图像处理】MATLAB：图像噪声

噪声模型

模拟噪声的行为和影响的能力是图像复原的核心。

采用函数imnoise来使用噪声污染一幅图像。

g = imnoise ( f , type , parameters )

噪声应用

高斯噪声在诸如低照明水平图像传感器操作等情况下被用作一种近似。
椒盐噪声由不完善的开关设备产生。
照相乳剂中的银粒大小是用对数正态分布描述的随机变量。
瑞利噪声产生于波段成像。
指数和厄兰噪声在描述激光成像中的噪声方面是很有用的。

使用指定的分布产生空间随机噪声（函数imnoise2）

该函数生成一个大小为M×N的噪声数组R，不以任何形式缩放。imnoise输出一个有噪声的图像，而imnoise2产生噪声模式本身。

function R = imnoise2(type, M, N, a, b)
%IMNOISE2 Generates an array of random numbers with specified PDF.
%   R = IMNOISE2(TYPE, M, N, A, B) generates an array, R, of size
%   M-by-N, whose elements are random numbers of the specified TYPE
%   with parameters A and B. If only TYPE is included in the
%   input argument list, a  single random number of the specified
%   TYPE and default parameters shown below is generated. If only
%   TYPE, M, and N are provided, the default parameters shown below
%   are used.  If M = N = 1, IMNOISE2 generates a single random
%   number of the specified TYPE and parameters A and B.
%
%   Valid values for TYPE and parameters A and B are:
%
%   'uniform'       Uniform random numbers in the interval (A, B).
%                   The default values are (0, 1).
%   'gaussian'      Gaussian random numbers with mean A and standard
%                   deviation B. The default values are A = 0, B = 1.
%   'salt & pepper' Salt and pepper numbers of amplitude 0 with
%                   probability Pa = A, and amplitude 1 with
%                   probability Pb = B. The default values are Pa =
%                   Pb = A = B = 0.05.  Note that the noise has
%                   values 0 (with probability Pa = A) and 1 (with
%                   probability Pb = B), so scaling is necessary if
%                   values other than 0 and 1 are required. The noise
%                   matrix R is assigned three values. If R(x, y) =
%                   0, the noise at (x, y) is pepper (black).  If
%                   R(x, y) = 1, the noise at (x, y) is salt
%                   (white). If R(x, y) = 0.5, there is no noise
%                   assigned to coordinates (x, y).
%   'lognormal'     Lognormal numbers with offset A and shape
%                   parameter B. The defaults are A = 1 and B =
%                   0.25.
%   'rayleigh'      Rayleigh noise with parameters A and B. The
%                   default values are A = 0 and B = 1.
%   'exponential'   Exponential random numbers with parameter A.  The
%                   default is A = 1.
%   'erlang'        Erlang (gamma) random numbers with parameters A
%                   and B.  B must be a positive integer. The
%                   defaults are A = 2 and B = 5.  Erlang random
%                   numbers are approximated as the sum of B
%                   exponential random numbers.%   Copyright 2002-2004 R. C. Gonzalez, R. E. Woods, & S. L. Eddins
%   Digital Image Processing Using MATLAB, Prentice-Hall, 2004
%   $Revision: 1.5 $  $Date: 2003/10/12 23:37:29 $% Set default values.
if nargin == 1a = 0; b = 1;M = 1; N = 1;
elseif nargin == 3a = 0; b = 1;
end% Begin processing. Use lower(type) to protect against input being
% capitalized.
switch lower(type)
case 'uniform'R = a + (b - a)*rand(M, N);
case 'gaussian'R = a + b*randn(M, N);
case 'salt & pepper'if nargin <= 3a = 0.05; b = 0.05;end% Check to make sure that Pa + Pb is not > 1.if (a + b) > 1error('The sum Pa + Pb must not exceed 1.')endR(1:M, 1:N) = 0.5;% Generate an M-by-N array of uniformly-distributed random numbers% in the range (0, 1). Then, Pa*(M*N) of them will have values <=% a. The coordinates of these points we call 0 (pepper% noise). Similarly, Pb*(M*N) points will have values in the range% > a & <= (a + b).  These we call 1 (salt noise). X = rand(M, N);c = find(X <= a);R(c) = 0;u = a + b;c = find(X > a & X <= u);R(c) = 1;
case 'lognormal'if nargin <= 3a = 1; b = 0.25;endR = a*exp(b*randn(M, N));
case 'rayleigh'R = a + (-b*log(1 - rand(M, N))).^0.5;
case 'exponential'if nargin <= 3a = 1;endif a <= 0error('Parameter a must be positive for exponential type.')endk = -1/a;R = k*log(1 - rand(M, N));
case 'erlang'if nargin <= 3a = 2; b = 5;endif (b ~= round(b) | b <= 0)error('Param b must be a positive integer for Erlang.')endk = -1/a;R = zeros(M, N);for j = 1:b         R = R + k*log(1 - rand(M, N));end
otherwiseerror('Unknown distribution type.')
end

代码示例

r = imnoise2('gaussian',100000,1,0,1);
hist(r,50);

运行结果

周期噪声（函数imnoise3）

function [r, R, S] = imnoise3(M, N, C, A, B)
%IMNOISE3 Generates periodic noise.
%    [r, R, S] = IMNOISE3(M, N, C, A, B), generates a spatial
%    sinusoidal noise pattern, r, of size M-by-N, its Fourier
%    transform, R, and spectrum, S.  The remaining parameters are:
%
%    C is a K-by-2 matrix with K pairs of freq. domain coordinates (u,
%    v) that define the locations of impulses in the freq. domain. The
%    locations are with respect to the frequency rectangle center at
%    (floor(M/2) + 1, floor(N/2) + 1).  The impulse locations are spe-
%    cified as increments with respect to the center. For ex, if M =
%    N = 512, then the center is at (257, 257). To specify an impulse
%    at (280, 300) we specify the pair (23, 43); i.e., 257 + 23 = 280,
%    and 257 + 43 = 300. Only one pair of coordinates is required for
%    each impulse.  The conjugate pairs are generated automatically.
%
%    A is a 1-by-K vector that contains the amplitude of each of the
%    K impulse pairs. If A is not included in the argument, the
%    default used is A = ONES(1, K).  B is then automatically set to
%    its default values (see next paragraph).  The value specified
%    for A(j) is associated with the coordinates in C(j, 1:2).
%
%    B is a K-by-2 matrix containing the Bx and By phase components
%    for each impulse pair.  The default values for B are B(1:K, 1:2)
%    = 0.%   Copyright 2002-2004 R. C. Gonzalez, R. E. Woods, & S. L. Eddins
%   Digital Image Processing Using MATLAB, Prentice-Hall, 2004
%   $Revision: 1.5 $  $Date: 2004/11/04 22:32:42 $% Process input parameters.
[K, n] = size(C);
if nargin == 3A(1:K) = 1.0;B(1:K, 1:2) = 0;
elseif nargin == 4B(1:K, 1:2) = 0;
end% Generate R.
R = zeros(M, N);
for j = 1:Ku1 = M/2 + 1 + C(j, 1); v1 = N/2 + 1 + C(j, 2);R(u1, v1) = i * (A(j)/2) * exp(i*2*pi*C(j, 1) * B(j, 1)/M);% Complex conjugate.u2 = M/2 + 1 - C(j, 1); v2 = N/2 + 1 - C(j, 2);R(u2, v2) = -i * (A(j)/2) * exp(i*2*pi*C(j, 2) * B(j, 2)/N);
end% Compute spectrum and spatial sinusoidal pattern.
S = abs(R);
r = real(ifft2(ifftshift(R)));

代码示例

C = [0 64;0 128;32 32;64 0;128 0;-32 32];
[r,R,S] = imnoise3(512,512,C);
subplot(1,2,1);imshow(S,[ ]);title('指定冲击的频谱');
subplot(1,2,2);imshow(r,[ ]);title('相应的正弦噪声模式');

运行结果

估计噪声参数

function [p, npix] = histroi(f, c, r)
%    用于计算一幅图像在多边形区域内的直方图，多边形的顶点由c和r指定，通过函数roipoly复制所定义的多边形区域% Generate the binary mask image.
B = roipoly(f, c, r);% Compute the histogram of the pixels in the ROI.
p = imhist(f(B));% Obtain the number of pixels in the ROI if requested in the output.
if nargout > 1npix = sum(B(:));
end

代码示例

f = imread('noisy_image.tif');
figure,imshow(f);
[B,c,r] = roipoly(f);
[p,npix] = histroi(f,c,r);
figure,bar(p,1);
X = imnoise2('gaussian',npix,1,147,20);
figure,hist(X,130);
axis([0 300 0 140]);

运行结果

（运行出了些问题…所以贴了书上的图…待以后回来修改，也许…）

【图像处理】MATLAB：图像噪声相关推荐

【图像处理】图像噪声去除实验用 matlab 实现KNN（K近邻平滑滤波器）和 SNN（对称近邻平滑滤波器）
实验内容实现KNN(K近邻平滑滤波器) 算法实现为函数[im]=KNN_denoise (I,K,N),其中I为读入的图像矩阵:K为最近邻个数,N为模板大小参数(N*N). 测试代码 im=imre ...
Matlab图像噪声处理
一．内容对一幅灰度图像f(x,y): (1)对f(x,y)加高斯白噪声和椒盐噪声: (2)分别画出原图和加噪后的图像及其各自对应的直方图: (3)用几何均值滤波分别对加高斯噪声和椒盐噪声图进行滤波处 ...
matlab 图像读取长宽_MATLAB图像处理基本操作
本文中对于大多数的操作,是对数字图像处理领域中最为著名的"lena"图片进行操作的.原图如下(Figure 1): 本文中对于大多数的操作,是对数字图像处理领域中最为著名的&quo ...
图像处理：图像中噪声分布和概率密度函数的关系
学习笔记:数字图像处理--图像中噪声分布和概率密度函数的关系学习的时候,看到各种噪声,高斯.锐利.伽马.均匀等.每个分布有概率密度函数,而这个和噪声有啥具体的连接的关系,我不是很懂,稍微研究了一下, ...
matlab 图像幅度谱低通滤波_数字图像处理期末复习2018-12-21
数字图像处理期末复习2018-12-21 愉快先生 0.204 · 字数 5547 · 阅读 1834 2018-12-22 19:35 (数字图像冈萨雷斯第二版教材) 一.基本原理图像的读取.存储 ...
MATLAB--数字图像处理图像噪声与滤波处理
一.实验名称图像的噪声与滤波处理二.实验目的 1.熟悉MATLAB软件的使用. 2.掌握图像的噪声与滤波处理. 三.实验内容 1.对一张图片添加不同强度的高斯噪声.椒盐噪声 2.对图像进行不同模板 ...
图像处理技术之三：降噪处理（中值、均值、最大值、最小值滤波、图像噪声）
图像噪声的成因分类与常见图像去噪算法简介 1.图像噪声的成因图像在生成和传输过程中常常因受到各种噪声的干扰和影响而是图像降质,这对后续图像的处理和图像视觉效应将产生不利影响.噪声种类很多,比如:电噪 ...
Matlab中消除图像噪声之中值滤波器：medfilt2
Matlab中消除图像噪声之中值滤波器:medfilt2 medfilt2, matlab, 消除噪声, 中值滤波器, 椒盐噪声在图像处理中,在进行如边缘检测这样的进一步处理之前,通常需要首先进行一 ...
9.2.1 Python图像处理之图像数学形态学-二值形态学应用之噪声消除
9.2.1 Python图像处理之图像数学形态学-二值形态学应用之噪声消除文章目录 9.2.1 Python图像处理之图像数学形态学-二值形态学应用之噪声消除 1 算法原理 2 代码 3 效果 1 ...
matlab图像类论文,基于matlab图形图像处理技术毕设论文.doc
基于matlab图形图像处理技术毕设论文基于MATLAB图形图像处理技术摘要本文提出了一种基于MATLAB的数字图像处理技术的设计,系统中包括了图像处理技术的各个方面,涵盖了图像处理领域的个别 ...

【图像处理】MATLAB：图像噪声

噪声模型

噪声应用

相关函数

使用指定的分布产生空间随机噪声（函数imnoise2）

周期噪声（函数imnoise3）

估计噪声参数

【图像处理】MATLAB：图像噪声相关推荐

最新文章

热门文章