how to make a cuda image processing filter in c-CodePudding

I am trying to make a image filter in Cuda c and i don't think i understand exactly how does the thread assignment for each pixel work.This is my code so far: imageFilter.cu

#include "cuda_runtime.h"
#include "device_launch_parameters.h"
#include "Filtering_Functions.h"

__global__ void Grayscale_image(int h, int w, unsigned char* Image) {
    int x = (blockIdx.x * blockDim.x)   threadIdx.x;
    int y = (blockIdx.y * blockDim.y)   threadIdx.y;
    unsigned int tid = threadIdx.y * blockDim.y   threadIdx.x;

    if (x > 0 && x < w - 1 && y > 0 && y < h - 1)
    {
        Image[tid] = 0.299 * Image[tid]   0.587 * Image[tid   1]   0.114 * Image[tid   2];
        Image[tid   1] = Image[tid];
        Image[tid   2] = Image[tid];
    }
}
void Image_Grayscale(unsigned char* Image, int Height, int Width) {
    unsigned char* Uploaded_Image = NULL;
    dim3 blocks(Width / 16, Height / 16);
    dim3 threads(16, 16);

    cudaMalloc((void**)&Uploaded_Image, Height * Width * 3);

    cudaMemcpy(Uploaded_Image, Image, Height * Width * 3, cudaMemcpyHostToDevice);
    
    Grayscale_image << <blocks, threads >> > (Height, Width, Uploaded_Image);

    cudaMemcpy(Image, Uploaded_Image, Height * Width * 3, cudaMemcpyDeviceToHost);

    cudaFree(Uploaded_Image);
}

filtering_functions.h

#ifndef filtering_functions
#define filtering_functions
void Image_Grayscale(unsigned char* Image, int Height, int Width);
#endif

ImageFiltering.cpp

#include <iostream>
#include <stdio.h>
#include <opencv2/opencv.hpp>
#include <opencv2/core/core.hpp>
#include <opencv2/highgui/highgui.hpp>
#include <opencv2/imgcodecs.hpp>
#include "Filtering_Functions.h"

using namespace std;
using namespace cv;

int main() {
    Mat Image = imread("Example.png");

    cout << "The uploaded image has Height: " << Image.rows << ", Width: " << Image.cols << endl;

    Image_Grayscale(Image.data, Image.rows, Image.cols);
    imwrite("Grayscale_Filter.png", Image);

    system("pause");
    return 0;
}

At the end i do not see any changes. Can someone tell me what am i doing wrong, or at least what i don't understand?

CodePudding user response：

Cuda operations are asynchonous relative to the host processor. I suspect you're attempting to write the data to disk before it's been touched by the GPU. Consider calling cudaStreamSynchronize(0) before trying to examine the results.

Cuda Stream Management

CodePudding user response：

The code you provided does not compile - you used an undeclared identifier Image2 in this line:

imwrite("Grayscale_Filter.png", Image2);

Maybe that's the issue? Are you writing something else to the file?