OpenCV 帶參數的維納濾波 C++

下圖是OpenCV 自帶例子的修改版本。

結果：

代碼實現：

#include <iostream>
#include "opencv2/imgproc.hpp"
#include "opencv2/imgcodecs.hpp"
#include <opencv2/opencv.hpp>

using namespace cv;
using namespace std;

void calcPSF(Mat& outputImg, Size filterSize, int len, double theta);
void fftshift(const Mat& inputImg, Mat& outputImg);
void filter2DFreq(const Mat& inputImg, Mat& outputImg, const Mat& H);
void calcWnrFilter(const Mat& input_h_PSF, Mat& output_G, double nsr);
void edgetaper(const Mat& inputImg, Mat& outputImg, double gamma = 5.0, double beta = 0.2);

int LEN = 50;
int THETA = 360;
int snr = 8000;
Mat imgIn;
Rect roi;
static void onChange(int pos, void* userInput);

int main(int argc, char *argv[])
{
    string strInFileName = "529g.tif";

    imgIn = imread(strInFileName, IMREAD_GRAYSCALE);
    if (imgIn.empty()) //check whether the image is loaded or not
    {
        cout << "ERROR : Image cannot be loaded..!!" << endl;
        return -1;
    }
    imshow( "src", imgIn );

    // it needs to process even image only
    roi = Rect(0, 0, imgIn.cols & -2, imgIn.rows & -2);
    imgIn = imgIn(roi);
    cv::namedWindow("inverse");

    createTrackbar("LEN", "inverse", &LEN, 200, onChange, &imgIn);
    onChange(0, 0);
    createTrackbar("THETA", "inverse", &THETA, 360, onChange, &imgIn);
    onChange(0, 0);
    createTrackbar("snr", "inverse", &snr, 10000, onChange, &imgIn);
    onChange(0, 0);
    imshow( "inverse", imgIn );
    cv::waitKey(0);

    return 0;
}

void calcPSF(Mat& outputImg, Size filterSize, int len, double theta)
{
    Mat h(filterSize, CV_32F, Scalar(0));
    Point point(filterSize.width / 2, filterSize.height / 2);
    ellipse(h, point, Size(0, cvRound(float(len) / 2.0)), 90.0 - theta,
            0, 360, Scalar(255), FILLED);
    Scalar summa = sum(h);
    outputImg = h / summa[0];
    Mat tmp;
    normalize(outputImg, tmp, 1,0, CV_MINMAX);
    imshow( "psf", tmp);
}
void fftshift(const Mat& inputImg, Mat& outputImg)
{
    outputImg = inputImg.clone();
    int cx = outputImg.cols / 2;
    int cy = outputImg.rows / 2;
    Mat q0(outputImg, Rect(0, 0, cx, cy));
    Mat q1(outputImg, Rect(cx, 0, cx, cy));
    Mat q2(outputImg, Rect(0, cy, cx, cy));
    Mat q3(outputImg, Rect(cx, cy, cx, cy));
    Mat tmp;
    q0.copyTo(tmp);
    q3.copyTo(q0);
    tmp.copyTo(q3);
    q1.copyTo(tmp);
    q2.copyTo(q1);
    tmp.copyTo(q2);
}
void filter2DFreq(const Mat& inputImg, Mat& outputImg, const Mat& H)
{
    Mat planes[2] = { Mat_<float>(inputImg.clone()), Mat::zeros(inputImg.size(), CV_32F) };
    Mat complexI;
    merge(planes, 2, complexI);
    dft(complexI, complexI, DFT_SCALE);
    Mat planesH[2] = { Mat_<float>(H.clone()), Mat::zeros(H.size(), CV_32F) };
    Mat complexH;
    merge(planesH, 2, complexH);
    Mat complexIH;
    mulSpectrums(complexI, complexH, complexIH, 0);
    idft(complexIH, complexIH);
    split(complexIH, planes);
    outputImg = planes[0];
}
void calcWnrFilter(const Mat& input_h_PSF, Mat& output_G, double nsr)
{
    Mat h_PSF_shifted;
    fftshift(input_h_PSF, h_PSF_shifted);
    Mat planes[2] = { Mat_<float>(h_PSF_shifted.clone()), Mat::zeros(h_PSF_shifted.size(), CV_32F) };
    Mat complexI;
    merge(planes, 2, complexI);
    dft(complexI, complexI);
    split(complexI, planes);
    Mat denom;
    pow(abs(planes[0]), 2, denom);
    denom += nsr;
    divide(planes[0], denom, output_G);
}
void edgetaper(const Mat& inputImg, Mat& outputImg, double gamma, double beta)
{
    int Nx = inputImg.cols;
    int Ny = inputImg.rows;
    Mat w1(1, Nx, CV_32F, Scalar(0));
    Mat w2(Ny, 1, CV_32F, Scalar(0));
    float* p1 = w1.ptr<float>(0);
    float* p2 = w2.ptr<float>(0);
    float dx = float(2.0 * CV_PI / Nx);
    float x = float(-CV_PI);
    for (int i = 0; i < Nx; i++)
    {
        p1[i] = float(0.5 * (tanh((x + gamma / 2) / beta) - tanh((x - gamma / 2) / beta)));
        x += dx;
    }
    float dy = float(2.0 * CV_PI / Ny);
    float y = float(-CV_PI);
    for (int i = 0; i < Ny; i++)
    {
        p2[i] = float(0.5 * (tanh((y + gamma / 2) / beta) - tanh((y - gamma / 2) / beta)));
        y += dy;
    }
    Mat w = w2 * w1;
    multiply(inputImg, w, outputImg);
}

// Trackbar call back function
static void onChange(int , void* userInput)
{
    Mat imgOut;
    //Hw calculation (start)
    Mat Hw, h;
    calcPSF(h, roi.size(), LEN, (double)THETA);
    calcWnrFilter(h, Hw, 1.0 / double(snr));
    //Hw calculation (stop)
    imgIn.convertTo(imgIn, CV_32F);
    edgetaper(imgIn, imgIn);
    // filtering (start)
    filter2DFreq(imgIn(roi), imgOut, Hw);
    // filtering (stop)
    imgOut.convertTo(imgOut, CV_8U);
    normalize(imgOut, imgOut, 0, 255, NORM_MINMAX);
//    imwrite("result.jpg", imgOut);
    imshow("inverse", imgOut);
}