IplImage像素访问JavaCV

Question

我正在尝试访问IplImage的一个像素一个像素。我正在使用Java和Processing，有时我需要逐个像素地访问。到目前为止，我已经做到了这一点，但我不知道哪里出了问题：

public IplImage PImageToIplImage(PImage imageSrc)
    {
        IplImage imageDst;
        if(imageSrc.format==RGB)
        {
            imageDst = IplImage.create(imageSrc.width, imageSrc.height, IPL_DEPTH_8U, 3);
            ByteBuffer imagePixels=imageDst.getByteBuffer();
            int locPImage, locIplImage, x, y;
            for(y=0; y<imageSrc.height; y++)
                for(x=0; x<imageSrc.width; x++)
                {
                    locPImage = x + y * width;
                    locIplImage=y*imageDst.widthStep()+3*x;
                    imagePixels.put(locIplImage+2, (byte)(red(imageSrc.pixels[locPImage])));
                    imagePixels.put(locIplImage+1, (byte)(green(imageSrc.pixels[locPImage])));
                    imagePixels.put(locIplImage, (byte)(blue(imageSrc.pixels[locPImage])));
                }
        }
}

在Karlphilip建议后，我来到了这个，仍然不工作。当我尝试显示时，它给了我一个nullPointer异常：

imageDst = IplImage.create(imageSrc.width, imageSrc.height, IPL_DEPTH_8U, 3);
CvMat imagePixels = CvMat.createHeader(imageDst.height(), imageDst.width(), CV_32FC1);  
cvGetMat(imageDst, imagePixels, null, 0); 
int locPImage, x, y;
for(y=0; y<imageSrc.height; y++)
   for(x=0; x<imageSrc.width; x++)
   {
       locPImage = x + y * width;
       CvScalar scalar = new CvScalar();
       scalar.setVal(0, red(imageSrc.pixels[locPImage]));
       scalar.setVal(1, green(imageSrc.pixels[locPImage]));
       scalar.setVal(2, blue(imageSrc.pixels[locPImage]));
       cvSet2D(imagePixels, y, x, scalar);
   }
   imageDst = new IplImage(imagePixels); 

Answer 1

在JavaCV中迭代每个像素的最快方法是：

ByteBuffer buffer = image.getByteBuffer();

for(int y = 0; y < image.height(); y++) {
    for(int x = 0; x < image.width(); x++) {
        int index = y * image.widthStep() + x * image.nChannels();

        // Used to read the pixel value - the 0xFF is needed to cast from
        // an unsigned byte to an int.
        int value = buffer.get(index) & 0xFF;

        // Sets the pixel to a value (greyscale).
        buffer.put(index, value);

        // Sets the pixel to a value (RGB, stored in BGR order).
        buffer.put(index, blue);
        buffer.put(index + 1, green);
        buffer.put(index + 2, red);
    }
}

Answer 2

下面的代码从磁盘加载图像，并通过迭代图像的像素来执行灰度转换：

    IplImage image = cvLoadImage("pipeline.png", CV_LOAD_IMAGE_COLOR);
    if (image == null) 
    {
        System.out.println("Erro ao carregar imagem!");
        return;
    }

    double r, g, b;
    r = g = b = 0.0;       

    CvMat mtx = CvMat.createHeader(image.height(), image.width(), CV_32FC1);  
    cvGetMat(image, mtx, null, 0); 

    System.out.println(mtx.rows() + "x" + mtx.cols());

    for (int i = 0; i < mtx.rows(); i++)
    {
        for (int j = 0; j < mtx.cols(); j++)
        {
            CvScalar rgb = cvGet2D(mtx, i, j);
            r = rgb.val(0);
            g = rgb.val(2);
            b = rgb.val(1);

            double gray = (r + g + b) / 3;

            CvScalar scalar = new CvScalar();
            scalar.setVal(0, gray);
            scalar.setVal(1, gray);
            scalar.setVal(2, gray);
            cvSet2D(mtx, i, j, scalar);
        }
    }        

    IplImage result = new IplImage(mtx); 
    cvSaveImage("manual_gray.png", result);

    cvReleaseImage(image); 

我不确定创建一个新的CvMat是否是最好的方法，但这是我所知道的在javacv中工作的唯一方法。

编辑

不幸的是，您使用的数据类型不是来自OpenCV、like PImage，但我已经尽了最大努力模拟您正在做的事情。

    // Create a black image
    IplImage imageDst = IplImage.create(imageSrc.width(), imageSrc.height(), IPL_DEPTH_8U, 3); 

    // Create a temporary mat to iterate over it's pixels
    CvMat imagePixels = CvMat.createHeader(imageDst.height(), imageDst.width(), CV_32FC1); 

    // Copy black image to temporary mat
    cvGetMat(imageDst, imagePixels, null, 0);

    int x, y;
    for(y=0; y<imageSrc.height(); y++)
       for(x=0; x<imageSrc.width(); x++)
       {
            // At this point you tried to do: locPImage = x + y * width;
            // but I think you might have mistaken for: locPImage = y + x * width;

            //...
       }
       imageDst = new IplImage(imagePixels);