用C++添加高斯噪聲

Question

我想給高斯噪聲的輸入圖像添加高斯噪聲PDF（概率分佈函數），我寫這個代碼並檢查很多次，但輸出不正確，我很困惑！

int main() { 
    Mat Frame; 
    string address; 
    printf("Please Drag and Drop Your Image"); 
    cin >> address; 
    Frame = imread(address, CV_LOAD_IMAGE_GRAYSCALE); 
    int arrayOfIntensity[256] = { 0 }, intensity; 

    //NEW COUNT INTENSITY OF EVERY PIXEL 
    for (int i = 0; i < Frame.rows; i++) 
     for (int j = 0; j < Frame.cols; j++) 
      arrayOfIntensity[Frame.at<uchar>(i, j)]++; 

    //SUM OF INTENSITY 
    int sumOfintensity = 0; 
    for (int i = 0; i < Frame.rows; i++) 
     for (int j = 0; j < Frame.cols; j++) 
      sumOfintensity += Frame.at<uchar>(i, j); 
    //AVG OF INTENSITY 
    double avgOfintensity = sumOfintensity, varOfintensity = 0; 
    avgOfintensity /= Frame.rows*Frame.cols; 
    //VARIANCE OF INTENSITY 
    for (int i = 0; i < Frame.rows; i++) 
     for (int j = 0; j < Frame.cols; j++) 
      varOfintensity += pow(Frame.at<uchar>(i, j) - avgOfintensity, 2); 
    varOfintensity/= Frame.rows*Frame.cols; 
    //PROBABILITY 
    float probability[256] = { 0 }, intermediate[256] = { 0 }, factor, sumProb[256] = { 0 }, newSumProb[256] = { 0 }; 
    factor = (sqrt(6.28)*avgOfintensity); 
    cout << "factor :" << factor << endl; 
    factor = 1/factor; 
    cout << "new factor :" << factor << endl; 
    for (int i = 0; i < 256; i++) { 
     intermediate[i] =-1*(pow(i - avgOfintensity, 2))/(2 * pow(varOfintensity, 2)); 
     probability[i] = factor*(pow(2.718281, intermediate[i])); 
     //SUM OF PROBABILTY 
     if (i == 0) 
      sumProb[i] = probability[i]; 
     else 
      sumProb[i] = probability[i] + sumProb[i - 1]; 
    } 
    //INTO 0-1 RANGE 
    for (int i = 0; i < 256; i++) 
     newSumProb[i] = sumProb[i]/ sumProb[255]; 

    float finalProb[256] = { 0 }; 
    for (int i = 0; i < 256; i++) { 
     double random = (rand() % 10)/1000000.0 +(rand() % 10)/100000.0 +(rand() % 10)/10000.0 +(rand() % 10)/1000.0 +(rand() % 10)/100.0+ (rand() % 10)/10.0; 
     for (int j = 0; j < 256; j++) { 
      if (random<newSumProb[j]) { 
       finalProb[i] = newSumProb[j]; 
       break; 
      } 
     } 
    } 
    int max = 0; 
    for (int i = 0; i < 256; i++) 
     if (finalProb[max]<finalProb[i]) 
      max = i; 

    for (int i = 0; i < 256; i++) 
     finalProb[i] =(finalProb[i] * 256.0)/ finalProb[max]; 

    for (int i = 0; i < Frame.rows; i++) 
     for (int j = 0; j < Frame.cols; j++) 
      Frame.at<uchar>(i, j) = saturate_cast<uchar>(finalProb[Frame.at<uchar>(i, j)]); 

    imshow("Result", Frame); 
    waitKey(); 
} 

正確的輸出： tihs image is correct output with matlab 我的不正確的輸出： this image is incorrect output with my code

Answer 1

據我瞭解，您的代碼不高斯噪聲的圖像添加。
高斯噪聲由2個值定義：平均值和標準差。

添加高斯噪聲的圖像，裝置生成新的畫面即幀，其中像素強度的分佈遵循正態分佈的大小，然後加入這對要進行實際的圖像嘈雜。

我還沒有試圖完全理解你的代碼，但你在做什麼似乎是像素相關的，並且比必要的方式更復雜。

要生成一個像素遵循正態分佈的框架，可以使用中心極限定理，該定理指出在相同分佈之後求和n個獨立變量趨於正態分佈。

因此，您可以設置幀的每個像素是：

/* Generate a pixel with a random intensity that follows the normal distribution */ 
int n = 30; 
int sum = 0; 
for(int k = 0; k < n; ++k) 
    sum += rand() % 255; 
pixel_i_j = sum/n; 

/* 
* pixel_i_j at this point follows a normal distribution with 
* parameters : 
* - mean = 256/2 = 128 
* - std = std of uniform law between 0 - 256/sqrt(n) 
* 
* So adapt it to the normal law defined by my parameters */ 
pixel_i_j = ((pixel_i_j - mean)/std) * my_std + my_mean;