这是课程第五次实验，时间太紧了，不是自己写的，难受，有空再改。
要是还能正常上课，我保证上课不睡瞌睡。
之前写的图像傅里叶变换的代码
https://www.cnblogs.com/cyssmile/p/12555602.html
这次实验的代码
#include<opencv2/opencv.hpp>
#include <vector>
#include <iostream>
#include <stdio.h>
#include <cmath>
#include <string>
using namespace cv;
using namespace std;
Mat DFT(const Mat& image);
void genIPF(Mat& H);
Mat IDFT(const Mat& image);
Mat IPF_Filter(const Mat& image);
Mat BPF_Filter(const Mat& image);
//D0截至频率 t通过系数
static int D0, t;
static int unq_n;
int main(int argc, char** argv)
{
Mat image = imread("D:/images/test.jpg", 0);
if (!image.data)
{
printf("No image data \n");
return -1;
}
//D0截至频率 t通过系数
D0 = 0;
t = 2;
//截至频率
unq_n = 0;
Mat new_img;
new_img = IPF_Filter(image);
Mat bfp_img;
bfp_img = BPF_Filter(image);
namedWindow("Display img", 0);
namedWindow("Display 理想高通/低通滤波", 0);
namedWindow("Display 布特沃斯滤波", 0);
//string name = "new_img";
imshow("Display img", image);
//理想高通/低通滤波
imshow("Display 理想高通/低通滤波", new_img);
//布特沃斯滤波
imshow("Display 布特沃斯滤波", bfp_img);
//imwrite(name, new_img);
waitKey(0);
return 0;
}
//理想高通/低通滤波器模板
void genIPF(Mat& H)
{
int P = H.rows;
int Q = H.cols;
for (int u = 0; u < P; u++)
for (int v = 0; v < Q; v++)
H.at<float>(u, v) = sqrt(pow(u - P / 2, 2) + pow(v - Q / 2, 2)) > D0 ? abs(0 - t) : abs(1 - t);
}
//布特沃斯滤波器
void genBPF(Mat& I)
{
int P = I.rows;
int Q = I.cols;
for (int u = 0; u < P; ++u)
for (int v = 0; v < Q; ++v)
if (t == 0)
I.at<float>(u, v) = 1 / (1 + pow(sqrt(pow(u - P / 2, 2) + pow(v - Q / 2, 2)) / D0, 2 * unq_n));
else
I.at<float>(u, v) = 1 / (1 + pow(D0 / sqrt(pow(u - P / 2, 2) + pow(v - Q / 2, 2)), 2 * unq_n));
}
Mat DFT(const Mat& image) {
int M = image.rows;
int N = image.cols;
int P = 2 * M, Q = 2 * N;
//填充中心化图像
Mat fill_img(P, Q, CV_32FC1);
for (int i = 0; i < M; i++)
for (int j = 0; j < N; j++)
fill_img.at<float>(i, j) = (image.at<uchar>(i, j)) * pow(-1, i + j);
for (int i = M; i < P; i++)
for (int j = N; j < Q; j++)
fill_img.at<float>(i, j) = 0;
//实部（图像本身），虚部（全0),组合到一个Mat
//Mat planes[] = {Mat_<float>(fill_img), Mat::zeros(fill_img.size(), CV_32F)};
Mat combine(P, Q, CV_32FC2, Scalar(0, 0));
for (int i = 0; i < P; i++)
for (int j = 0; j < Q; j++)
combine.at<Vec2f>(i, j)[0] = fill_img.at<float>(i, j);
//对组合Mat进行dft
dft(combine, combine);
//对数变换，计算1+log|F|提高显示效果
Mat f_chart(P, Q, CV_32FC1);
for (int i = 0; i < P; i++)
for (int j = 0; j < Q; j++)
f_chart.at<float>(i, j) = 1 + log(sqrt(pow(combine.at<Vec2f>(i, j)[0], 2) + pow(combine.at<Vec2f>(i, j)[1], 2)));
f_chart.convertTo(f_chart, CV_8UC1);
normalize(f_chart, f_chart, 0, 255, NORM_MINMAX);
namedWindow("Display f_chart", WINDOW_NORMAL);
imshow("Display f_chart", f_chart);
imwrite("f_chart.jpg", f_chart);
return combine;
}
Mat IDFT(const Mat& image) {
int P = image.rows;
int Q = image.cols;
Mat f(P, Q, CV_32FC1);
dft(image, f, DFT_INVERSE | DFT_REAL_OUTPUT | DFT_SCALE);
for (int i = 0; i < P; i++)
for (int j = 0; j < Q; j++)
f.at<float>(i, j) *= pow(-1, i + j);
f.convertTo(f, CV_8UC1);
return f;
}
Mat IPF_Filter(const Mat& image)
{
Mat F = DFT(image);
//生成滤波器
int M = image.rows;
int N = image.cols;
int P = 2 * M, Q = 2 * N;
Mat H(P, Q, CV_32FC1);
genIPF(H);
//滤波
Mat G = F;
for (int i = 0; i < P; i++)
for (int j = 0; j < Q; j++) {
//虚实部都要滤波
G.at<Vec2f>(i, j)[0] = F.at<Vec2f>(i, j)[0] * H.at<float>(i, j);
G.at<Vec2f>(i, j)[1] = F.at<Vec2f>(i, j)[1] * H.at<float>(i, j);
}
//逆变换并截取
Mat new_img = IDFT(G)(Range(0, M), Range(0, N));
return new_img;
}
Mat BPF_Filter(const Mat& image)
{
Mat F = DFT(image);
int M = image.rows;
int N = image.cols;
int P = 2 * M, Q = 2 * N;
Mat H(P, Q, CV_32FC1);
genBPF(H);
Mat G = F;
for (int i = 0; i < P; i++)
for (int j = 0; j < Q; j++)
{
G.at<Vec2f>(i, j)[0] = F.at<Vec2f>(i, j)[0] * H.at<float>(i, j);
G.at<Vec2f>(i, j)[1] = F.at<Vec2f>(i, j)[1] * H.at<float>(i, j);
}
Mat new_img = IDFT(G)(Range(0, M), Range(0, N));
return new_img;
}
实验效果
