TBB、OpenMP、SSE、AVX加速sobel算子

TBB加速

#include <tbb\tbb.h>
double TBB_Sobel(Mat srcImage, int kernelSize, Mat &gradX, Mat &gradY){
	class GetGradient
	{
	public:
		Mat *gx, *gy;//x和y方向梯度图
		Mat src;
		void operator()(const tbb::blocked_range<int> &r) const{//这个方法保证本对象可以调用迭代器的参数
			int step = src.step;
			int stepxy = gx->step;
			uchar *psrcImage = src.data;
			uchar *px = gx->data;
			uchar *py = gy->data;
			for (int rindex = r.begin(); rindex != r.end(); ++rindex){//!=
				for (int j = 1; j < src.cols - 1; j++)
				{
					gx->ptr<short>(rindex)[j] = (psrcImage[((rindex - 1)*step + j + 1)] + psrcImage[(rindex*step + j + 1)] * 2 + psrcImage[((rindex + 1)*step + j + 1)] - psrcImage[((rindex - 1)*step + j - 1)] - psrcImage[(rindex*step + j - 1)] * 2 - psrcImage[((rindex + 1)*step + j - 1)]);
					gy->ptr<short>(rindex)[j] = (psrcImage[(rindex - 1)*step + j - 1] + psrcImage[(rindex - 1)*step + j] * 2 + psrcImage[(rindex - 1)*step + j + 1] - psrcImage[(rindex + 1)*step + j - 1] - psrcImage[(rindex + 1)*step + j] * 2 - psrcImage[(rindex + 1)*step + j + 1]);
					
				}
			}
			}
	};
	gradX = Mat::zeros(srcImage.size(), CV_16SC1);//定义相同尺寸的矩阵保存x方向梯度图
	gradY = Mat::zeros(srcImage.size(), CV_16SC1);//定义相同尺寸的矩阵保存y方向梯度图
	//遍历,跳过第一行第一列和最后一行最后一列
	//----------------------------------------------------------------------------------------//
	GetGradient m_GetGradient;
	m_GetGradient.gx = &gradX;
	m_GetGradient.gy = &gradY;
	m_GetGradient.src = srcImage;
	parallel_for(blocked_range<int>(1, srcImage.rows - 1), m_GetGradient, auto_partitioner());
	//---------------------------------------------------------------------------------------//记录算法运行时间
	//convertScaleAbs(gradX, gradX);
	//convertScaleAbs(gradY, gradY);//
	return 0;
}

OpenMP加速

#include "omp.h"
double OpenMP_Sobel(Mat srcImage, int kernelSize, Mat &gradX, Mat &gradY){
	if (srcImage.empty() || srcImage.channels() != 1)
	{
		return -1;
	}
	if (srcImage.co