OpenCVSharp学习笔记(2)--Mat基本图像容器

1、Mat类型

为解决原C语言中IplImage内存管理问题，引入C++类Mat自动管理内存

Mat类由两个部分组成：

1）矩阵头matrix header(包含矩阵尺寸、存储方法、存储地址等等），矩阵头的尺寸是常数

2）一个指向存储所有像素值的矩阵（所选存储方法的不同可以是不同的维数）的指针matrix pointer，

引用计数制管理Mat对象的内存

每个Mat对象有自己的矩阵头(matrix header)，但可能共享的相同图像矩阵(矩阵指针指向相同地址)

copy构造函数，只拷贝矩阵头和矩阵指针，不拷贝图像矩阵

如果要拷贝矩阵用clone或copyTo方法

//读取图像
using var lenaA = Cv2.ImRead(ImagePath.LenaColor, ImreadModes.Color);
if (lenaA.Empty()) return;

//共享图像矩阵
var lenaB = new Mat(lenaA, Range.All, Range.All);
using var lenaC = lenaA;

//获取指定区域图像
using var lenaD = new Mat(lenaA, new Rect(100, 250,370, 180));
using var lenaE = lenaA[new Range(100, 200), Range.All];

//复制图像矩阵
using var lenaF = lenaA.Clone();
using Mat lenaG = new Mat();
lenaA.CopyTo(lenaG);

//修改某个区域为绿
lenaA[new Rect(50, 100, 150, 200)] = new Mat(new Size(150, 200), MatType.CV_8UC3,Scalar.Green);

//显示所有图像
using (new Window("lenaA", lenaA))
using (new Window("lenaB", lenaB))
using (new Window("lenaC", lenaC))
using (new Window("lenaD", lenaD))
using (new Window("lenaE", lenaE))
using (new Window("lenaF", lenaF))
using (new Window("lenaG", lenaG))
{
    Cv2.WaitKey();
}

显示结果如下：

可见对lenaA矩阵的修改，同时影响了lenaB、lenaC、lenaD和lenaE,

通过Clone()与CopyTo()方法的lenaF、lenaG未受影响。

2、存储方法

通过指定颜色空间和数据类型存储像素值。

常见的颜色空间：灰度空间Gray、RGB空间、HSV等等

数据类型

[typeof(byte)] = MatType.CV_8UC1,
[typeof(sbyte)] = MatType.CV_8SC1,
[typeof(short)] = MatType.CV_16SC1,
[typeof(char)] = MatType.CV_16UC1,
[typeof(ushort)] = MatType.CV_16UC1,
[typeof(int)] = MatType.CV_32SC1,
[typeof(float)] = MatType.CV_32FC1,
[typeof(double)] = MatType.CV_64FC1,
[typeof(Vec2b)] = MatType.CV_8UC2,
[typeof(Vec3b)] = MatType.CV_8UC3,
[typeof(Vec4b)] = MatType.CV_8UC4,
[typeof(Vec6b)] = MatType.CV_8UC(6),
[typeof(Vec2s)] = MatType.CV_16SC2,
[typeof(Vec3s)] = MatType.CV_16SC3,
[typeof(Vec4s)] = MatType.CV_16SC4,
[typeof(Vec6s)] = MatType.CV_16SC(6),
[typeof(Vec2w)] = MatType.CV_16UC2,
[typeof(Vec3w)] = MatType.CV_16UC3,
[typeof(Vec4w)] = MatType.CV_16UC4,
[typeof(Vec6w)] = MatType.CV_16UC(6),
[typeof(Vec2i)] = MatType.CV_32SC2,
[typeof(Vec3i)] = MatType.CV_32SC3,
[typeof(Vec4i)] = MatType.CV_32SC4,
[typeof(Vec6i)] = MatType.CV_32SC(6),
[typeof(Vec2f)] = MatType.CV_32FC2,
[typeof(Vec3f)] = MatType.CV_32FC3,
[typeof(Vec4f)] = MatType.CV_32FC4,
[typeof(Vec6f)] = MatType.CV_32FC(6),
[typeof(Vec2d)] = MatType.CV_64FC2,
[typeof(Vec3d)] = MatType.CV_64FC3,
[typeof(Vec4d)] = MatType.CV_64FC4,
[typeof(Vec6d)] = MatType.CV_64FC(6),
[typeof(Point)] = MatType.CV_32SC2,
[typeof(Point2f)] = MatType.CV_32FC2,
[typeof(Point2d)] = MatType.CV_64FC2,
[typeof(Point3i)] = MatType.CV_32SC3,
[typeof(Point3f)] = MatType.CV_32FC3,
[typeof(Point3d)] = MatType.CV_64FC3,
[typeof(Size)] = MatType.CV_32SC2,
[typeof(Size2f)] = MatType.CV_32FC2,
[typeof(Size2d)] = MatType.CV_64FC2,
[typeof(Rect)] = MatType.CV_32SC4,
[typeof(Rect2f)] = MatType.CV_32FC4,
[typeof(Rect2d)] = MatType.CV_64FC4,
[typeof(DMatch)] = MatType.CV_32FC4

MatType类型

格式：CV_[数据类型]C[通道数]

如：MatType.CV_8UC3 表示3通道8位无符号整型

MatType.CV_64FC4 表示4通道64位浮点型

3、创建Mat对象

//定义3x2,8位无符号整型3通道的矩阵
using var M = new Mat(3, 2, MatType.CV_8UC3, new Scalar(0, 0, 255));
Debug.WriteLine(M);
Debug.WriteLine("M=\n" + M.Dump());

//重新调整矩阵大小，内容随机
M.Create(4, 3, MatType.CV_8UC2);
Debug.WriteLine("after Create M=\n" + M.Dump());

//MATLAB形式的初始化方式
using var eye = (Mat)Mat.Eye(4, 4, MatType.CV_64F);
Debug.WriteLine("eye=\n" + eye.Dump());

using var ones = (Mat)Mat.Ones(new Size(2, 2), MatType.CV_32F);
Debug.WriteLine("ones=\n" + ones.Dump());

using var zeros = (Mat)Mat.Zeros(3, 3, MatType.CV_8UC1);
Debug.WriteLine("zeros=\n" + zeros.Dump());

//小矩阵赋予初始值
var C = new Mat<Double>(3, 3, new Double[] { 0, -1, 0, -1, 5, -1, 0, -1, 0 });
Debug.WriteLine("C=\n" + C.Dump());

4、格式化输出

//格式化输出
Debug.WriteLine("Default:\n" + C.Dump(FormatType.Default));
Debug.WriteLine("MATLAB:\n" + C.Dump(FormatType.MATLAB));
Debug.WriteLine("CSV:\n" + C.Dump(FormatType.CSV));
Debug.WriteLine("Python:\n" + C.Dump(FormatType.Python));
Debug.WriteLine("NumPy:\n" + C.Dump(FormatType.NumPy));
Debug.WriteLine("C:\n" + C.Dump(FormatType.C));

源码下载：LearnOpenCVSharp

参考链接：

Mat - 基本图像容器

Mat - The Basic Image Container