分别采用GPU、CPU对图像进行sobel滤波处理

#include <stdio.h>
#include "cuda_runtime.h"
#include "device_launch_parameters.h"
#include<math.h>
#include <malloc.h> 
#include <opencv2/opencv.hpp>#include <stdlib.h>#define BLOCK_SIZE 1//图像卷积 GPU
__global__ void sobel_gpu(unsigned char* in, unsigned char* out, const int Height, const int Width)
{int x = blockDim.x * blockIdx.x + threadIdx.x;int y = blockDim.y + blockIdx.y + threadIdx.y;int index = y * Width + x;int Gx = 0;int Gy = 0;unsigned char x0, x1, x2, x3, x4, x5, x6, x7, x8;if (x>0 && x<(Width-1) && y>0 && y<(Height-1)){x0 = in[(y - 1)*Width + (x - 1)];x1 = in[(y - 1)*Width + (x)];x2 = in[(y - 1)*Width + (x + 1)];x3 = in[(y)*Width + (x - 1)];x5 = in[(y)*Width + (x + 1)];x6 = in[(y + 1)*Width + (x - 1)];x7 = in[(y + 1)*Width + (x)];x8 = in[(y + 1)*Width + (x + 1)];Gx = (x0 + 2 * x3 + x6) - (x2 + 2 * x5 + x8);Gy = (x0 + 2 * x1 + x2) - (x6 + 2 * x7 + x8);out[index] = (abs(Gx) + abs(Gy)) / 2;}
}//Sobel滤波  CPU实现
void sobel_cpu(cv::Mat srcImg, cv::Mat dstImg, int Height, int Width)
{int Gx = 0;int Gy = 0;for (int i = 1; i < Height - 1; i++){unsigned char* dataUp = srcImg.ptr<unsigned char>(i - 1);unsigned char* data = srcImg.ptr<unsigned char>(i);unsigned char* dataDown = srcImg.ptr<unsigned char>(i + 1);unsigned char* out = dstImg.ptr<unsigned char>(i);for (int j = 1; j < Width - 1; j++){Gx = (dataUp[j + 1] + 2 * data[j + 1] + dataDown[j + 1]) - (dataUp[j - 1] + 2 * data[j - 1] + dataDown[j - 1]);Gy = (dataUp[j - 1] + 2 * dataUp[j] + dataUp[j + 1]) - (dataDown[j - 1] + 2 * dataDown[j] + dataDown[j + 1]);out[j] = (abs(Gx) + abs(Gy)) / 2;}}
}int main()
{cv::Mat src;src = cv::imread("photo16.jpg");cv::Mat grayImg,gaussImg;cv::cvtColor(src, grayImg, cv::COLOR_BGR2GRAY);cv::GaussianBlur(grayImg, gaussImg, cv::Size(3,3), 0, 0, cv::BORDER_DEFAULT);int height = src.rows;int width = src.cols;//输出图像cv::Mat dst_gpu(height, width, CV_8UC1, cv::Scalar(0));//GPU存储空间int memsize = height * width * sizeof(unsigned char);//输入 输出unsigned char* in_gpu;unsigned char* out_gpu;cudaMalloc((void**)&in_gpu, memsize);cudaMalloc((void**)&out_gpu, memsize);dim3 threadsPreBlock(BLOCK_SIZE, BLOCK_SIZE);dim3 blocksPreGrid((width + threadsPreBlock.x - 1)/threadsPreBlock.x, (height + threadsPreBlock.y - 1)/threadsPreBlock.y);cudaMemcpy(in_gpu, gaussImg.data, memsize, cudaMemcpyHostToDevice);sobel_gpu <<<blocksPreGrid, threadsPreBlock>>> (in_gpu, out_gpu, height, width);cudaMemcpy(dst_gpu.data, out_gpu, memsize, cudaMemcpyDeviceToHost);//cudaDeviceSynchronize();//输出图像cv::Mat dst_cpu(height, width, CV_8UC1, cv::Scalar(0));sobel_cpu(gaussImg, dst_cpu, height, width);cv::imwrite("dst_cpu_save.png", dst_cpu);cv::imwrite("dst_gpu_save.png", dst_gpu);//cv::namedWindow("src", cv::WINDOW_NORMAL);cv::imshow("src", src);//cv::namedWindow("dst_cpu", cv::WINDOW_NORMAL);cv::imshow("dst_cpu", dst_cpu);//cv::namedWindow("dst_gpu", cv::WINDOW_NORMAL);cv::imshow("dst_gpu", dst_gpu);cv::waitKey();cudaFree(in_gpu);cudaFree(out_gpu);return 0;
}