C++图像处理之双边滤波

本文为大家分享了C++图像处理之双边滤波的具体代码，供大家参考，具体内容如下

1、 近期在学习双边滤波相关知识，其原理如下（以后补上）：

2 、灰度图双边滤波实现C++代码如下，网上大多数是基于8位灰度图和彩色图像的。（此次代码未经优化，可去除opencv依赖）：

//灰度图双边滤波
void m_bilateralFilter(cv::Mat src,cv::Mat& dst,int radius,float sigma_r,float sigma_d)
{
if (src.empty())
return;
if (dst.empty())
{
dst = src.clone();
}
if (src.depth() == CV_16U){
for (int i = radius; i < src.rows - radius; i++)
 for (int j = radius; j < src.cols - radius; j++)
 {
  float sum_1 = .0f, sum_2 = .0f;
  for (int k = 0; k < 2 * radius - 1; k++)
   for (int l = 0; l < 2 * radius - 1; l++)
   {
    int dis_x = radius - k;
    int dis_y = radius - l;
    int coord_x_image = i - radius + k;
    int coord_y_image = j - radius + l;
    float dis_spatial = dis_x*dis_x + dis_y*dis_y;
    float dis_range = (src.at<unsigned short>(i, j)
     - src.at<unsigned short>(coord_x_image, coord_y_image))*(src.at<unsigned short>(i, j)
     - src.at<unsigned short>(coord_x_image, coord_y_image));
    float c_tmp = exp(-dis_spatial /
     (2 * sigma_d * sigma_d));
    float s_tmp = exp(-dis_range /
     (2 * sigma_r * sigma_r));
    sum_1 += c_tmp*s_tmp*src.at<unsigned short>(coord_x_image, coord_y_image);
    sum_2 += c_tmp*s_tmp;

   }
  dst.at<unsigned short>(i, j) = sum_1 / sum_2;
 }
}
else if (src.depth() == CV_8U)
{
for (int i = radius; i < src.rows - radius; i++)
 for (int j = radius; j < src.cols - radius; j++)
 {
  float sum_1 = .0f, sum_2 = .0f;
  for (int k = 0; k < 2 * radius - 1; k++)
   for (int l = 0; l < 2 * radius - 1; l++)
   {
    int dis_x = radius - k;
    int dis_y = radius - l;
    int coord_x_image = i - radius + k;
    int coord_y_image = j - radius + l;
    float dis_spatial = dis_x*dis_x + dis_y*dis_y;
    float dis_range = (src.at<unsigned char>(i, j)
     - src.at<unsigned char>(coord_x_image, coord_y_image))*(src.at<unsigned char>(i, j)
     - src.at<unsigned char>(coord_x_image, coord_y_image));
    float c_tmp = exp(-dis_spatial /
     (2 * sigma_d * sigma_d));
    float s_tmp = exp(-dis_range /
     (2 * sigma_r * sigma_r));
    sum_1 += c_tmp*s_tmp*src.at<unsigned char>(coord_x_image, coord_y_image);
    sum_2 += c_tmp*s_tmp;

   }
  dst.at<unsigned char>(i, j) = sum_1 / sum_2;
 }
}
}

3、目前是基于单通道图像，效果如下：

原图：

opencv 库的效果（cv::bilateralFilter(img_src, img_dst, 10,10 * 2, 10 / 2)）

该程序的效果（m_bilateralFilter(img_src, img_dst, 5, 10 * 2, 10 / 2)）

以上就是本文的全部内容，希望对大家的学习有所帮助，也希望大家多多支持服务器之家。

原文链接：https://blog.csdn.net/Asimov_Liu/article/details/79861649