基于16位图像的LEE、KUAN图像去噪

double **matToDouble(Mat src)//将图像转化为二维数组
{
	int row = src.rows;
	int col = src.cols;
	double **a;
	a = new double*[row];
	for (int i = 0; i < row; i++)
	{
		a[i] = new double[col];
	}

	for (int i = 0; i < src.rows; i++)
		for (int j = 0; j < src.cols; j++)
		{
			a[i][j] = (double)src.at < ushort >(i, j);
		}
	return a;
}

void KUAN(Mat &src, Mat &dst)
{
	double **i1, **i2;
	i1 = matToDouble(src);
	double ybar;
	double ystad;
	double ENL;
	double sx2;
	double xcap;
	for (int i = 1; i < src.rows - 1; ++i)
		for (int j = 1; j < src.cols - 1; ++j)
		{
			//ybar = (i1[i][j] + i1(i - 1, j - 1) + i1(i - 1, j) + i1(i, j - 1)
			//+ i1(i + 1, j) + i1(i, j + 1) + i1(i + 1, j + 1) + i1(i - 1, j + 1) + i1(i + 1, j - 1)) / 9;
			ybar = (i1[i - 1][j - 1] + i1[i - 1][j] + i1[i - 1][j + 1] + i1[i][j - 1] + i1[i][j] + i1[i][j + 1] + i1[i + 1][j - 1] + i1[i + 1][j] + i1[i + 1][j + 1]) / 9;
			/*ystad = sqrt(((src.at<double>(i, j) - ybar)*(src.at<double>(i, j) - ybar)
			+ (src.at<double>(i - 1, j - 1) - ybar)*(src.at<double>(i - 1, j - 1) - ybar)
			+ (src.at<double>(i - 1, j) - ybar)*(src.at<double>(i - 1, j) - ybar)
			+ (src.at<double>(i, j - 1) - ybar)*(src.at<double>(i, j - 1) - ybar)
			+ (src.at<double>(i + 1, j) - ybar)*(src.at<double>(i + 1, j) - ybar)
			+ (src.at<double>(i, j + 1) - ybar)*(src.at<double>(i, j + 1) - ybar)
			+ (src.at<double>(i + 1, j + 1) - ybar)*(src.at<double>(i + 1, j + 1) - ybar)
			+ (src.at<double>(i - 1, j + 1) - ybar)*(src.at<double>(i - 1, j + 1) - ybar)
			+ (src.at<double>(i + 1, j - 1) - ybar)*(src.at<double>(i + 1, j - 1) - ybar)) / 9);*/
			ystad = sqrt(((i1[i][j] - ybar)*(i1[i][j] - ybar) + (i1[i - 1][j - 1] - ybar)*(i1[i - 1][j - 1] - ybar)
				+ (i1[i - 1][j] - ybar)*(i1[i - 1][j] - ybar) + (i1[i][j + 1] - ybar)*(i1[i][j + 1] - ybar)
				+ (i1[i][j - 1] - ybar)*(i1[i][j - 1] - ybar) + (i1[i][j + 1] - ybar)*(i1[i][j + 1] - ybar)
				+ (i1[i + 1][j - 1] - ybar)*(i1[i + 1][j - 1] - ybar) + (i1[i + 1][j] - ybar)*(i1[i + 1][j] - ybar)
				+ (i1[i + 1][j + 1] - ybar)*(i1[i + 1][j + 1] - ybar)) / 9);

			ENL = (ybar / (ystad + 0.1))*(ybar / (ystad + 0.1));
			sx2 = (ENL*ystad*ystad - ybar*ybar) / (ENL + 1);
			xcap = ybar + (sx2*(i1[i][j] - ybar)) / (sx2 + ((ybar*ybar + sx2) / ENL) + 0.1);
			if (xcap>16384)
			{
				xcap = 16384;
			}
			else if (xcap<0)
			{
				xcap = 0;
			}
			else
			{
				xcap = (int)(xcap + 0.5);
			}
			dst.at<ushort>(i, j) = xcap;

		}

}


void Lee(Mat &src, Mat &dst)
{
	double **i1, **i2;
	i1 = matToDouble(src);
	double ybar;
	double ystad;
	double ENL;
	double sx2;
	double xcap;
	for (int i = 1; i < src.rows - 1; ++i)
		for (int j = 1; j < src.cols - 1; ++j)
		{
			ybar = (i1[i - 1][j - 1] + i1[i - 1][j] + i1[i - 1][j + 1] + i1[i][j - 1] + i1[i][j] + i1[i][j + 1] + i1[i + 1][j - 1] + i1[i + 1][j] + i1[i + 1][j + 1]) / 9;
			ystad = sqrt(((i1[i][j] - ybar)*(i1[i][j] - ybar) + (i1[i - 1][j - 1] - ybar)*(i1[i - 1][j - 1] - ybar)
				+ (i1[i - 1][j] - ybar)*(i1[i - 1][j] - ybar) + (i1[i][j + 1] - ybar)*(i1[i][j + 1] - ybar)
				+ (i1[i][j - 1] - ybar)*(i1[i][j - 1] - ybar) + (i1[i][j + 1] - ybar)*(i1[i][j + 1] - ybar)
				+ (i1[i + 1][j - 1] - ybar)*(i1[i + 1][j - 1] - ybar) + (i1[i + 1][j] - ybar)*(i1[i + 1][j] - ybar)
				+ (i1[i + 1][j + 1] - ybar)*(i1[i + 1][j + 1] - ybar)) / 9);

			ENL = (ybar / (ystad + 0.1))*(ybar / (ystad + 0.1));
			sx2 = (ENL*ystad*ystad - ybar*ybar) / (ENL + 1);
			xcap = ybar + (sx2*(i1[i][j] - ybar)) / (sx2 + (ybar*ybar / ENL) + 0.1);
			if (xcap > 16384)
			{
				xcap = 16384;
			}
			else if (xcap < 0)
			{
				xcap = 0;
			}
			else
			{
				xcap = (int)(xcap + 0.5);
			}
			dst.at<ushort>(i, j) = xcap;

		}

}

对于8位图像，把代码中的ushort改成uchar即可，下面是读取16位raw数据到mat的代码

int RawToMat(const char fileName[], cv::Mat& dst,const int height,const int width)
{
	if (fileName == nullptr) return 0;
	if ((width <= 0) | (height <= 0)) return 0;

	size_t nsize = width * height;

	unsigned short *data = new unsigned short[nsize];

	FILE *file;

	file = fopen(fileName, "rb+");

	fread(data, sizeof(unsigned short), nsize, file);

	fclose(file);

	cv::Mat temp(height, width, CV_16UC1, data);

	dst = temp.clone();

	return 0;
}

int saveRawFile(const char fileName[], cv::Mat src)
{
	if (fileName == nullptr) return 0;

	FILE *fw = fopen(fileName, "wb");

	fwrite(src.data, src.elemSize(), src.cols * src.rows, fw);

	fclose(fw);

	return 1;
}

调用测试：

int main()
{
	
	Mat src;
	char strPath[256] = "2020/4.raw";
	int h = 1024;//1024;
	int w = 1024;
	RawToMat(strPath, src, h, w);

	Mat out = src.clone();
	KUAN(src, out);
	img.saveRawFile("2020/out.raw", out);//16位数据
	waitKey();
	return 0;

}

速度比 opencv 的GaussianBlur慢点，cv库的高斯函数也能处理16位图。