工具： MATLAB；

动机：使用 MATLAB toolbox 程序设计相对简单，通过初步学习一些图像处理的常见技术，帮助建立这方面的技能体系；

面向对象：有过经验，但是一头雾水的我。

 

 

1 图像、视频文件读取

 

clear all; close all; clc;% Read an imageA = imread('lena.jpg');% Display the read imagefigure, imshow(A);% Print height, width and number of channels of the read imageheight = size(A, 1);width = size(A, 2);number_of_channels = size(A, 3);% Resize the image to 2x and displayB = imresize(A, 2.0);figure, imshow(B);% Rotate the image 逆时针旋转C = imrotate(A, 90);figure, imshow(C);

　关键点：imread,figure(imshow),size获取高度，宽度，图像通道数的参数，图像缩放，旋转。

clear all; close all; clc;% Read an imageA = imread('lena.jpg');% Translate the imageA_trans = imtranslate(A, [5 15]);% Write the transformed image to diskimwrite(A_trans, 'newlena.jpg');figure,imshow(A_trans);

　关键：imwrite及其调用前做的工作。

1-1 视频读取（选看）

clear all; close all; clc;% Initialize an object for reading video videoObj = vision.VideoFileReader('video1.mp4', 'ImageColorSpace', 'RGB'); % Get video frame size and frame rate S = info(videoObj);width = S.VideoSize(1);height = S.VideoSize(end);frame_rate = S.VideoFrameRate;% Get individual video framesimage_data = step(videoObj);% Run the loop until all frames have been displayedwhile ~isDone(videoObj)    % Display video frame one by one    imshow(image_data);        image_data = step(videoObj);end% Release the object for reading videorelease(videoObj);

　关键：step，其流程reader读取生成对象，得到信息，逐帧显示 , PS:视频教程中有另外一种实现。 

3 图像格式

 3.1 RGB、灰度图片

clear all; close all; clc;% Read a RGB imageA = imread('lena.jpg');% Verify number of channelsnumber_of_channels = size(A, 3)% Convert RGB image to GrayscaleA_gray = rgb2gray(A);number_of_channels2 = size(A_gray, 3)figure, subplot(1, 2, 1), imshow(A), title('Input RGB image');subplot(1, 2, 2), imshow(A_gray), title('Converted Grayscale image');

　

关键：灰度图像颜色通道位1，rgb图像颜色通道数是3，subplot 显示多个图像，便于查看。

clear all; close all; clc;% Read a RGB imageA = imread('lena.jpg');% Extract individual R, G & B channels from RGB imageR = A(:, :, 1);G = A(:, :, 2);B = A(:, :, 3);% Convert RGB image to HSV color spaceA_hsv = rgb2hsv(A);% Extract individual H, S & V channels from HSV imageH = A_hsv(:, :, 1);S = A_hsv(:, :, 2);V = A_hsv(:, :, 3);% Display the read RGB image and HSV converted imagefigure, subplot(1, 2, 1), imshow(A), title('Read RGB image');subplot(1, 2, 2), imshow(A_hsv), title('Converted HSV image');% Displaying individual image channels.figure, subplot(1, 3, 1), imshow(S), title('Saturation channel');subplot(1, 3, 2), imshow(H), title('Hue channel');subplot(1, 3, 3), imshow(V), title('Value channel');

　　

关键：R、G、B、H、S、V 都是提取空间域（矩阵表示）的信息，位图片大小，RGB、HSV格式都是3个颜色通道，前者unit8，后者double格式。PS：HSV 有用的是V（当前）。

4 直方图及其均衡化

 直方图以前都是空间域上对像素操作，直方图则是借助频率来获取图像的统计信息 imhist。

clear all; close all; clc;% Read an imageA = imread('lena.jpg');B = imread('lena.jpg');% Convert the read image to single-channelA = rgb2gray(A);% Plot and display the histogram if A is a single channel image with% default number of binsfigure, subplot(1,3,1),imhist(A), title('Histogram of read image');        subplot(1,3,2),imshow(A),title('Grayscale image');        subplot(1,3,3),imshow(B),title('RGB image');

　　

直方图：左边暗，右边亮，不同亮度出现的频率，注意直方图应该表示一个颜色通道的统计信息，不要把RGB、HSV等格式图片多个通道的格式混合起来，这样是得不到可分析的信息的，RGB如下直方图生成方式如下：

clear all; close all; clc;% Read an imageA = imread('lena.jpg');% Plot and display the histogram if A is a 3 channel image with% default number of binsfigure, imhist(A(:, :, 1)), title('Histogram of 1st channel of read image');figure, imhist(A(:, :, 2)), title('Histogram of 2nd channel of read image');figure, imhist(A(:, :, 3)), title('Histogram of 3rd channel of read image');

　　

 

 5 直方图均衡化

亮度分布集中在左边图片太暗，集中在右边图片太亮，直方图均衡化就是让图像的不同像素点亮度均匀话，对灰度图片和其他格式图片有不同的做法

clear all; close all; clc;% Read an imageA = imread('test1.jpg');% Convert the image to single channel imageA = rgb2gray(A);% Equalize the histogram of read single channel imageA_histeq = histeq(A);% Display original and equalized images side by side with their respective% histogram plotsfigure, subplot(2, 2, 1), imshow(A), title('Original Image');subplot(2, 2, 2), imshow(A_histeq), title('Equalized Image');subplot(2, 2, 3), imhist(A), title('Histogram of Original Image');subplot(2, 2, 4), imhist(A_histeq), title('Histogram of Equalized Image');

　　

关键：imhist,histeq,

clear all; close all; clc;% Read a RGB imageA = imread('test1.jpg');% Convert the RGB image into HSV imageA_hsv = rgb2hsv(A);% Perform histogram equalization of V-channel in HSV image which is the 3rd% channelA_hsv(:, :, 3) = histeq(A_hsv(:, :, 3));% Convert back equalized HSV image into RGB imageA_histeq = hsv2rgb(A_hsv);% Display original and equalized RGB images side by sidefigure, subplot(1, 2, 1), imshow(A), title('Original RGB image');subplot(1, 2, 2), imshow(A_histeq), title('Equalized RGB image');

　　

关键：histeq(A(:,:,3) ,A是hsv格式，流程：RGB图片转化为HSV 图片，然后对HSV图片的V进行均衡化，均衡化后重新转化为RGB图片。

 6 图像平滑化（image smoothing）

释义：消除、弱化、压缩图像的噪声、边缘、细节、突变。可以在空间域（滤波）和频域上进行。fspecial、infilter

空间域： 生成一个任何大小的矩阵，用该矩阵对图像某个像素及其周围像素进行集体操作后代替原像素。根据将像素值设置为不同的值类型，可以分为：最小化，最大化，均值，加权均值操作。

空间滤波示例，流程：生成灰度图像，生成滤波器，应用滤波，执行滤波操作，PS：中值滤波过程不一样。

clear all; close all; clc;% Read an input imageA = imread('test1.jpg');% Conversion to single channel imageA = rgb2gray(A);% Generate Average filteraverage_h = fspecial('average', [9 9]);% Generate Gaussian filtergaussian_h = fspecial('gaussian', [13 13]);% Filter input image using average filterB_average = imfilter(A, average_h);% Filter input image using Gaussian filterB_gaussian = imfilter(A, gaussian_h);% Filter input image using median filterB_median = medfilt2(A, [5 5]);% Display input and filtered images side by sidefigure, subplot(2, 2, 1), imshow(A), title('Original input image');subplot(2, 2, 2), imshow(B_average), title('Image filtered using Averaging filter');subplot(2, 2, 3), imshow(B_gaussian), title('Image filtered using Gaussian filter');subplot(2, 2, 4), imshow(B_median), title('Image filtered using Median filter');

　　

均值滤波是最简单的空间滤波，特点：1、非加权；2、适用于去噪声，但会使图片模糊；3、突出总体细节

6.2 中值滤波（加权平滑滤波器）

pixel 有不同权值的滤波器，特点：越靠近中心的像素越重要；与非权值的滤波操作（均值滤波）相比，更好的模糊化，滤波器构造如下图所示。

补充一个导向滤波的例子，这么做的目的是保留图像的边缘。

clear all; close all; clc;% Read an imageA = imread('test1.jpg');% Filter the image using guided filterA_guided_filtered = imguidedfilter(A);% Display original and processed image side-by-sideimshowpair(A, A_guided_filtered, 'montage');

　　

7 边缘检测

边缘是重要的图像强度局部变化像素点的集合。边缘方向：垂直于图像强度最大强度变化的方向。

思路：空间域上对 x, y 求导，大于特定的值就是边缘，不同的算子生成方式有不同的近似求导方案。

clear all; close all; clc;% Read an imageA = imread('test1.jpg');% Convert the image to single-channel intensity image if the read image% is RGB (3-channel)A_gray = rgb2gray(A);% Method for Sobel edge detectionBW_sobel = edge(A_gray, 'sobel');% Method for Prewitt edge detectionBW_prewitt = edge(A_gray, 'prewitt');% Method of Canny edge detectionBW_canny = edge(A_gray, 'canny');% Display the images togetherfigure, subplot(2, 2, 1), imshow(A_gray), title('Original');subplot(2, 2, 2), imshow(BW_sobel), title('Sobel');subplot(2, 2, 3), imshow(BW_prewitt), title('Prewitt');subplot(2, 2, 4), imshow(BW_canny), title('Canny');

 

 

8 二值图像(im2bw)

图像只有两个值0,1,根据直方图左边暗，右边亮，左边取0（黑），右边取1（白），二值图像是根据直方图，大于某个值则设置为1（白色），二值图像生成的关键就是这个阈值的选择，否则二值图像生成的背景、人物、物体会杂糅在一起，没有办法继续处理。

流程：灰度 - double - im2bw，第2个参数要 ∈[0,1] ，阈值除以 255 带入 。 PS:  otsu 算法已经集成在matlab 工具箱中了

 

clear all;close all;clc; % Read an input imageA = imread('test1.jpg');% Convert the image to single-channel grayscale imageA_gray = rgb2gray(A);figure,imhist(A_gray),title('hist of A_grey');% Convert image to double i.e., [0,1]A_gray = im2double(A_gray);% Generate threhold value using Otsu's algorithmotsu_level = graythresh(A_gray);% Threshold image using Otsu's threshold and manually defined% threshold valuesB_otsu_thresh = im2bw(A_gray, otsu_level);B_thresh_50 = im2bw(A_gray, 50/255);B_thresh_100 = im2bw(A_gray, 100/255);B_thresh_150 = im2bw(A_gray, 150/255);B_thresh_200 = im2bw(A_gray, 200/255);% Display original and thresholded binary images side-by-sidefigure, subplot(2, 3, 1), imshow(A_gray), title('Original image');subplot(2, 3, 2), imshow(B_otsu_thresh), title('Binary image using Otsu threshold value');subplot(2, 3, 3), imshow(B_thresh_50), title('Binary image using threshold value = 50');subplot(2, 3, 4), imshow(B_thresh_100), title('Binary image using threshold value = 100');subplot(2, 3, 5), imshow(B_thresh_150), title('Binary image using threshold value = 150');subplot(2, 3, 6), imshow(B_thresh_200), title('Binary image using threshold value = 200');

　

9 图像降噪 （ fsepcial,imfilter）

噪声来源于 图像采集及转化、传输过程。

噪声分类： 高斯：正态分布；椒盐：一定概率分布的黑白像素点；冲击噪声：随机白点。

降噪手段：最简单的是均值滤波，但是会模糊整个图片，中值滤波：适用于椒盐噪声。

周期性的噪声多来自于电子电磁干扰，噪声在图像中的分布具有规律性（周期性），傅里叶域的频域技术咋消除周期性噪声是最有效的。带阻滤波：适合移除图片噪声及移除特定范围的频率。

clear all; close all; clc;% Read an input imageA = imread('test1.jpg');% Generate Gaussian and Average filtersh_gaussian = fspecial('gaussian');h_average = fspecial('average');% Filter input image using generated Gaussian and Average filtersB_gaussian = imfilter(A, h_gaussian);B_average = imfilter(A, h_average);% Filter input image using Median filter% Median filtering function takes as input grayscale imageA_gray = rgb2gray(A);B_median = medfilt2(A_gray);% Display original and filtered images side-by-side for comparing the% result of image de-noisingfigure, subplot(2, 2, 1), imshow(A), title('Original image with Salt & Pepper noise');subplot(2, 2, 2), imshow(B_gaussian), title('Input image filtered using Gaussian filter');subplot(2, 2, 3), imshow(B_average), title('Input image filtered using Average filter');subplot(2, 2, 4), imshow(B_median), title('Input image filtered using Median filter');

　　

 

 10 形态学

Erosion 腐蚀 ：imerode 分离相连物体，去毛刺，让物体变小；Dilation 膨胀 imdilate，拼接物体，填表面小坑，物体变大。

 

close all; clear all; clc;% Read an input imageA = imread('binaryImg1.jpg');% Convert to single channelA = rgb2gray(A);% Generate structuring element for usese = strel('disk', 15); % Perform image erosionB_eroded = imerode(A, se);% Perform image dilationB_dilated = imdilate(A, se);% Display the images side-by-side for comparisonfigure, subplot(1, 3, 1), imshow(A), title('Original read image');subplot(1, 3, 2), imshow(B_eroded), title('Original image after Erosion');subplot(1, 3, 3), imshow(B_dilated), title('Original image after Dilation');

开运算：先腐蚀再膨胀，分离物体、去毛刺、去小点（物体点），去小部件，表面坑变小。

闭运算：先膨胀再腐蚀，表面坑被填，内部背景点消失，表面坑被填，毛刺不变。

 

clear all; close all; clc;% Read an input imageA = imread('binaryImg1.jpg');% Convert read image to single channel imageA = rgb2gray(A);% Generate structuring element for processingse = strel('disk', 15); % Perform image opening operationB_open = imopen(A, se);% Perform image closing operationB_close = imclose(A, se);% Display original and processed imagesfigure, subplot(1, 3, 1), imshow(A), title('Original image');subplot(1, 3, 2), imshow(B_open), title('Original image after Opening');subplot(1, 3, 3), imshow(B_close), title('Original image after Closing');

 

 

 

　　

 

    

 

 

 利用膨胀得到物体边缘

clear all; close all; clc;% Read an input imageA = imread('binaryImg1.jpg');% Convert the read image to single channel imageA = rgb2gray(A);% Generate structuring element for usese = strel('disk', 5); % Perform image dilation% To get the object boundary, subtract the original image from the dilated% version of the orginal imageB_dilated = imdilate(A, se);% Subtract the original image from dilated imageB_boundary = B_dilated - A;% Display images side by sidefigure, subplot(1, 3, 1), imshow(A), title('Original image');subplot(1, 3, 2), imshow(B_dilated), title('Original image after Dilation');subplot(1, 3, 3), imshow(B_boundary), title('Original image with highlighted binary object boundaries');

　　

利用腐蚀得到物体边缘

clear all; close all; clc;% Read an input imageA = imread('binaryImg1.jpg');% Convert the read image to single channel imageA = rgb2gray(A);% Generate structuring element for usese = strel('disk', 5); % Perform image erosionB_eroded = imerode(A, se);% Subtract the eroded image from original imageB_boundary = A - B_eroded;% Display images side by sidefigure, subplot(1, 3, 1), imshow(A), title('Original image');subplot(1, 3, 2), imshow(B_eroded), title('Original image after Erosion');subplot(1, 3, 3), imshow(B_boundary), title('Original image with highlighted binary object boundaries');