OpenCV imwrite保存完整的黑jpeg。

I have done some pre processing for dft , and i am trying to save this image by imwrite.

我已经为dft做了一些预处理，我试图通过imwrite来保存这个图像。

My cropped image has this information

我裁剪的图像有这个信息。

output.type()           5   

output.channels()       1

output.depth()          5

But everytime when i save it gives black output. I have checked old existing threads of * but all seems not working for me. e.g. OpenCV2.3 imwrite saves black image

但是每次我保存它的时候都会输出黑色。我检查了旧的*的线程，但是似乎都不适合我。OpenCV2.3 imwrite可以保存黑色图像。

I have tried many color conversions and depth conversion as well but i dont know why it does not work.

我也尝试过许多颜色转换和深度转换，但我不知道为什么它不起作用。

std::vector<int> qualityType;
qualityType.push_back(CV_IMWRITE_JPEG_QUALITY);
qualityType.push_back(90);

Mat out1,out2;

cv::cvtColor(output, out1, CV_GRAY2BGR);
//out1.convertTo(out2,CV_8U,1./256); // i tried this too 
cv::imwrite("dft1.jpg",out1,qualityType); // tried even using quality type

imshow display this image fine only problem comes when i save it.

imshow显示这个图像很好，只有当我保存它的时候才会出现问题。

please help

请帮助

[EDIT] Maybe my dft class that i made has problem because i whenever i use dft function then the output can only work with inshow but for save it does not work.

也许我的dft类有问题，因为每当我使用dft函数时，输出只能与inshow一起工作，但对于保存它不起作用。

CDftRidgeAnalyses::CDftRidgeAnalyses(void)
{
}

CDftRidgeAnalyses::~CDftRidgeAnalyses(void)
{
}


Mat CDftRidgeAnalyses::GetRidgeAnalyses(Mat inpGray)
{
    Mat img = inpGray;
    int WidthPadded=0,HeightPadded=0;
    WidthPadded=img.cols*2;
    HeightPadded=img.rows*2;
    int M = getOptimalDFTSize( img.rows );
    //Create a Gaussian Highpass filter 5% the height of the Fourier transform
    double db  = 0.05 * HeightPadded;

    Mat fft = ForierTransform(img.clone(),HeightPadded,WidthPadded);

    Mat ghpf = CreateGaussianHighPassFilter(Size(WidthPadded, HeightPadded), db);
    Mat res;

    cv::mulSpectrums(fft,ghpf,res,DFT_COMPLEX_OUTPUT);

    Mat mag  = GetDftToImage(res,img);

    int cx = mag.cols/2;
    int cy = mag.rows/2;
    cv::Mat croped = mag(cv::Rect(0,0,cx, cy));

    cv::threshold(mag, mag, 0.019, 1, cv::THRESH_BINARY);

    Mat bgr;
    cvtColor(mag,bgr,CV_GRAY2RGB);

    //imshow("XXX",bgr);
    //imshow("croped", croped);
    //imshow("img",img);
    //
    //cv::waitKey();
    return croped;
}




Mat CDftRidgeAnalyses::ForierTransform(Mat inpGray,int M,int N)
{
    Mat img = inpGray;
    int i = img.channels();

    Mat padded;
    Mat img2;
    img.convertTo(img2,CV_64F,1./255);

    copyMakeBorder(img2, padded, 0, M - img2.rows, 0, N - img2.cols, BORDER_CONSTANT, Scalar::all(0));

    Mat element1 = Mat_<float>(padded);
    Mat element2 = Mat::zeros(padded.size(), CV_32F);
    Mat planes[] = {element1, element2};
    Mat complexImg;

    merge(planes, 2, complexImg);

    dft(complexImg, complexImg ,0, img.rows);

    //printMat(complexImg);

    return complexImg;
}



double CDftRidgeAnalyses::pixelDistance(double u, double v)
{
    return cv::sqrt(u*u + v*v);
}

double CDftRidgeAnalyses::gaussianCoeff(double u, double v, double d0)
{
    double d = pixelDistance(u, v);
    return 1.0 - cv::exp((-d*d) / (2*d0*d0));
}

cv::Mat CDftRidgeAnalyses::CreateGaussianHighPassFilter(cv::Size size, double cutoffInPixels)
{
    Mat ghpf(size, CV_32F);

    cv::Point center2((size.width*0.80), size.width/2);
    //cv::Point center2(0,0);

    for(int u = 0; u < ghpf.rows; u++)
    {
        for(int v = 0; v < ghpf.cols; v++)
        {
            ghpf.at<float>(u, v) = gaussianCoeff(u - center2.x, v - center2.y, cutoffInPixels);
        }
    }

    Mat bmp;

    int channels = ghpf.channels();
    int type = ghpf.type();
    int depth = ghpf.depth();

    cv::cvtColor(ghpf,bmp,CV_GRAY2RGB); 
    cv::cvtColor(ghpf,bmp,CV_GRAY2BGRA); 
    //imshow("XXX",bmp);
    int cx = ghpf.cols/2;
    int cy = ghpf.rows/2;
    Mat tmp;
    int iExactright =  (size.width*0.59);
    int iExactbottom = (size.height*0.86);
    //full   Mat q0(ghpf, Rect(69,10,400,290));

//  Mat whiteq(ghpf, Rect(0,390,270,330));

    int iMainleft=0, iMainright=0;
    int iMainBottom=0,iMainTop=0;


    Mat Quad;
    Mat ql(ghpf, Rect(190,0,270,330));

    /** Make the rectangle on middle default filter with respect to top right angle**/
    iMainleft=(size.width*0.111);
    iMainright=(size.width*0.402);
    iMainTop=(size.height*0.484);
    iMainBottom = (size.height*0.155);
    Quad = ghpf(Rect(iMainleft,iMainTop,iMainright+6,iMainBottom));
    Mat qTopRight(ghpf, Rect(iExactright,0, iMainright+6, iMainBottom));
    Quad.copyTo(qTopRight);

    /** Make the rectangle on middle default filter with respect to top left angle**/
    iMainright=(size.width*0.402);
    Quad = ghpf(Rect(300,iMainTop,300,iMainBottom));
    Mat qTopLeft(ghpf, Rect(0,0, 300, iMainBottom));
    Quad.copyTo(qTopLeft);


    /** Make the rectangle on middle default filter with respect to bottom left angle**/
    iMainTop = iMainTop-iMainBottom;
    iExactbottom = size.height - iMainBottom;
    Quad = ghpf(Rect(300,iMainTop,300,iMainBottom));
    Mat qBottomLeft(ghpf, Rect(0,iExactbottom, 300, iMainBottom));
    Quad.copyTo(qBottomLeft);

    /** Make the rectangle on middle default filter with respect to bottom right angle**/
    iMainleft=(size.width*0.111);
    iMainright=(size.width*0.402);

    Quad = ghpf(Rect(iMainleft,iMainTop,iMainright+6,iMainBottom));
    Mat qBottomRight(ghpf, Rect(iExactright,iExactbottom, iMainright+6, iMainBottom));
    Quad.copyTo(qBottomRight);


    // remove middle rectangle [ circle ] 
    iMainright=(size.width*0.402);
    Quad = ghpf(Rect(0,iMainTop+iMainTop,size.width,iMainBottom+iMainBottom-130));
    Mat qMiddle(ghpf,Rect(0,iMainTop+150,size.width,iMainBottom+iMainBottom-130));
    Quad.copyTo(qMiddle);
    qMiddle =ghpf(Rect(0,iMainTop-10,size.width,iMainBottom+iMainBottom-130));
    Quad.copyTo(qMiddle);

    normalize(ghpf, ghpf, 0, 1, CV_MINMAX);

    /*Mat x;
    cv::resize(ghpf,x,cv::Size(400,700));
    imshow("fftXhighpass2", x);*/
    Filter = ghpf;
    Mat padded;

    copyMakeBorder(ghpf, padded, 0, size.height - ghpf.rows, 0, size.width - ghpf.cols, BORDER_CONSTANT, Scalar::all(0));

    Mat planes[] = {Mat_<float>(padded), Mat::zeros(padded.size(), CV_32F)};
    Mat complexImg;

    merge(planes, 2, complexImg);


    return complexImg;
}

Mat CDftRidgeAnalyses::GetDftToImage(Mat res,Mat orgImage)
{
    idft(res,res,DFT_COMPLEX_OUTPUT,orgImage.rows);

    Mat padded;
    copyMakeBorder(orgImage, padded, 0,orgImage.rows, 0, orgImage.cols, BORDER_CONSTANT, Scalar::all(0));
    Mat planes[] = {Mat_<float>(padded), Mat::zeros(padded.size(), CV_32F)};
    split(res, planes);
    magnitude(planes[0], planes[1], planes[0]);
    Mat mag = planes[0];
    mag += Scalar::all(1);
  //  log(mag, mag);

    // crop the spectrum, if it has an odd number of rows or columns
    mag = mag(Rect(0, 0, mag.cols & -2, mag.rows & -2));

    normalize(mag, mag, 1, 0, CV_MINMAX);

    return mag;
}

The output that i am trying to save is from

我试图保存的输出来自于。

Mat org = imread("4.png",CV_LOAD_IMAGE_GRAYSCALE);
Mat re;
resize(org,re,cv::Size(311,519));
Mat xyz = CDftRidgeAnalyses::GetRidgeAnalyses(re);
cv::imwrite("dft1.jpg",xyz);

here the matrix xyz has these values

这里的矩阵xyz有这些值。

output.type()           5   

output.channels()       1

output.depth()          5

I hope you guys now can help me more better... maybe after converion from complex filter i am losing some points????

我希望你们现在能帮我更好的……也许在从复杂的滤波器中转换后，我失去了一些点????

2 个解决方案

#1

This 'feels' like a problem with floating point numbers and integers. When your image has floating point values, the imshow() of opencv expects these values to be between 0 and 1:

这个“感觉”就像浮点数和整数的问题。当您的图像具有浮点值时，opencv的imshow()期望这些值在0到1之间:

http://opencv.itseez.com/modules/highgui/doc/user_interface.html?highlight=imshow#cv2.imshow

http://opencv.itseez.com/modules/highgui/doc/user_interface.html?highlight=imshow cv2.imshow

I'm not quite sure what imwrite() does with floating point images, as I could not read it here:

我不太清楚什么是imwrite()和浮点图像，因为我不能在这里读到:

http://opencv.itseez.com/modules/highgui/doc/reading_and_writing_images_and_video.html?highlight=imwrite#cv2.imwrite

http://opencv.itseez.com/modules/highgui/doc/reading_and_writing_images_and_video.html?highlight=imwrite cv2.imwrite

Anyhow, imwrite might expect integer values between 0 and 255, and might simply cast floats to integers. In this case, almost everything is casted to 0 (i.g. 0.8 is casted to 0), hence your black images.

无论如何，imwrite可能会期望在0到255之间的整数值，并且可能只是将浮点数转换成整数。在这种情况下，几乎所有东西都被转换为0 (ig . 0.8被casted到0)，因此您的黑色图像。

Try to convert your images to CV_U8CX. Alternatively, here is something I use to debug such opencv problems:

尝试将您的图像转换为CV_U8CX。或者，这里有一些我用来调试opencv问题的东西:

void printType(Mat &mat) {
         if(mat.depth() == CV_8U)  printf("unsigned char(%d)", mat.channels());
    else if(mat.depth() == CV_8S)  printf("signed char(%d)", mat.channels());
    else if(mat.depth() == CV_16U) printf("unsigned short(%d)", mat.channels());
    else if(mat.depth() == CV_16S) printf("signed short(%d)", mat.channels());
    else if(mat.depth() == CV_32S) printf("signed int(%d)", mat.channels());
    else if(mat.depth() == CV_32F) printf("float(%d)", mat.channels());
    else if(mat.depth() == CV_64F) printf("double(%d)", mat.channels());
    else                           printf("unknown(%d)", mat.channels());
}

void printInfo(const char *prefix, Mat &mat) {
    printf("%s: ", prefix);
    printf("dim(%d, %d)", mat.rows, mat.cols);
    printType(mat);
    printf("\n");
}

void printInfo(Mat &mat) {
    printf("dim(%d, %d)", mat.rows, mat.cols);
    printType(mat);
    printf("\n");
}

This way you can find out what your cv::Mat has in it's data-field.

这样你就能知道你的简历是什么了:Mat在它的数据领域。

PS: I did not debug your code throughly, so stay open to other probleem causes.

我没有彻底地调试你的代码，所以要对其他的原因保持开放的态度。

#2

imwrite prints on a 0 to 255 scale, but your image is in a 0 to 1 scale. To scale up, use this line:

imwrite在0到255之间，但是你的图像是0到1的比例。为了扩大规模，请使用这一行:

image.convertTo(image, CV_8UC3, 255.0);