如果我按照opencv论坛里面的傅立叶频域处理的话，对频域图像归一化后滤波，但是逆变化得到的确实一幅全黑的图像。求解应该怎么办。我最终想要得到平滑处理的图像，类似于matlab中fft2和ifft的。
下面是我的图像和程序
#include <iostream>
#include <cv.h>
#include <highgui.h>
#include "imgproc/imgproc.hpp"
using namespace std;
using namespace cv;
int main()
{
Mat img = imread("C:\\Users\\LPX\\Desktop\\lena.jpg",0);
imshow("Original Image",img);
Mat dftInput1, dftImage1, inverseDFT, inverseDFTconverted;
img.convertTo(dftInput1, CV_32F);
dft(dftInput1, dftImage1);
imshow("dft",dftImage1);
int cx = dftImage1.cols/2;
int cy = dftImage1.rows/2;
Mat q0(dftImage1, Rect(0,0,cx,cy)); // Top-Left
Mat q1(dftImage1, Rect(cx,0,cx,cy)); // Top-Right
Mat q2(dftImage1, Rect(0,cy,cx,cy)); // Bottom-Left
Mat q3(dftImage1, Rect(cx,cy,cx,cy)); // Bottom-Right
Mat tmp;
q0.copyTo(tmp);
q3.copyTo(q0);
tmp.copyTo(q3);
q1.copyTo(tmp);
q2.copyTo(q1);
tmp.copyTo(q2);
//normalize(dftImage1, dftImage1, 0, 1, CV_MINMAX);
imshow("***",dftImage1);
Scalar s0=sum(dftImage1);
int m=dftImage1.rows;
int n=dftImage1.cols;
Mat gaussianBlur(m,n, CV_32F);
float D0 = 2*50*50;
for(int i=0; i<m; i++)
{
for(int j=0; j<n; j++)
{
float d = (i-m/2)*(i-m/2) +(j-n/2)*(j-n/2);
float m=-d/D0;
gaussianBlur.at<float>(i,j)=exp(m);
}
}
multiply(dftImage1, gaussianBlur, gaussianBlur);
int cx1 = gaussianBlur.cols/2;
int cy1 = gaussianBlur.rows/2;
Mat r0(gaussianBlur, Rect(0,0,cx1,cy1)); // Top-Left
Mat r1(gaussianBlur, Rect(cx1,0,cx1,cy1)); // Top-Right
Mat r2(gaussianBlur, Rect(0,cy1,cx1,cy1)); // Bottom-Left
Mat r3(gaussianBlur, Rect(cx1,cy1,cx1,cy1)); // Bottom-Right
Mat tmp1;
r0.copyTo(tmp1);
r3.copyTo(q0);
tmp1.copyTo(q3);
r1.copyTo(tmp1);
r2.copyTo(r1);
tmp1.copyTo(r2);
imshow("gaussian",gaussianBlur);
idft(gaussianBlur, inverseDFT, DFT_SCALE | DFT_REAL_OUTPUT ); // Applying IDFT
inverseDFT.convertTo(inverseDFTconverted, CV_8U);
Scalar s1=sum(gaussianBlur);
imshow("Output", inverseDFTconverted);
cout<<s0.val[0]<<endl;
cout<<s1.val[0];
waitKey(0);
return 0;
}
