边缘检测：laplace

laplace算子是二阶导
众所周知
$$f^{\prime }(x)=\frac{f(x+\delta )-f(x)}{\delta }(\delta 趋于0的极限)$$
而与$\delta$趋于0的方式无关，可以大于0趋于0，可以小于0趋于0，甚至可以震荡趋于0，而图像是离散的，所以求导也只能近似代替
比如取$\delta =1或者-1$
即$f^{\prime }(x)=f(x+1)-f(x)$或者$f^{\prime }(x)=f(x)-f(x-1)$
此时二阶导$f^{\prime \prime }(x)=f^{\prime }(x+1)-f^{\prime }(x)=f(x+1)-f(x)-f(x)+f(x-1)$(第一次选择$\delta$大于0趋于0，后一个两个均选择小于0趋于0，至于为啥这样做，感觉是为了对称性，或者是同一个方向累积的话误差会变大比如$f^{\prime \prime }(x)=f^{\prime }(x+1)-f^{\prime }(x)=f(x+2)-f(x+1)-f(x+1)+f(x)$显然太偏了)

计算方式确定后，核就定了，计算就是卷积比较简单

#include<iostream>
#include<opencv2\opencv.hpp>

using namespace std;
using namespace cv;

void BGR2Gray(Mat src, Mat &grayPic)
{

	for (int i = 0; i < src.rows; i++)
	{
		for (int j = 0; j < src.cols; j++)
		{
			//Gray = R*0.299 + G*0.587 + B*0.114

			grayPic.at<uchar>(i, j) = src.at<Vec3b>(i, j)[0] * 0.114 + src.at<Vec3b>(i, j)[1] * 0.587 + src.at<Vec3b>(i, j)[2] * 0.299;
		}

	}
}

void mylaplace(int lp[][3], Mat src, Mat out)
{
	
	for (int i = 1; i < src.rows-1; i++)
	{
		for (int j = 1; j < src.cols-1; j++)
		{
			int tmp = 0;
			for (int i1 = -1; i1 <= 1; i1++)
			{
				for (int i2 = -1; i2 <= 1; i2++)
				{
					tmp += lp[1 + i1][1 + i2] * src.at<uchar>(i + i1, j + i2);
				}
			}
			if (tmp > 255)
				tmp = 255;
			if (tmp < 0)
				tmp = 0;
			out.at<uchar>(i, j) = tmp;

		}
	}
}



int main()
{
	int lp[3][3] = { 0, 1, 0,
				     1, -4, 0,
				     0, 1, 0 };
	int lp2[3][3]= { 1, 1, 1,
		             1, -8, 1,
		             1, 1, 1 };
	Mat pic = imread("C:\\Users\\Thinkpad\\Desktop\\c++work\\pic\\lena.jpg", 1);

	Mat src = Mat::zeros(pic.rows, pic.cols, CV_8UC1);
	Mat out = src.clone();

	BGR2Gray(pic, src);
	mylaplace(lp2, src, out);


	imshow("1", out);
	waitKey();


	return 0;
}