C++, Negative RGB values of pixels using OpenCV

Question

I\'m using OpenCV to iterate through an image and find the colour of each pixel, here\'s some of the code I\'m using:

IplImage* img = cvLoadImage(\"c:\\\\tes         


        

Answer 1

I don't understand why you are accessing pixel data blindly (by 1D index) while you already know the image dimension (proved by the usage of IplImage::height and IplImage::width) and opencv does provide function to access pixel value (cvGet*D).

You are getting odd values in the direct pointer access because you did not factor in the byte padding performed by OpenCV. You may find in 8 bit images, (width*nChannels <= widthStep) for this reason.

A quick revised code can be as follow:

IplImage* img = cvLoadImage("c:\\test.png");
for(int iRow=0; iRow<img->height; ++iRow)
{
    for(int iCol=0; iCol<img->width; ++iCol)
    {
        CvScalar pixelVal = cvGet2D( img, iRow, iCol);
        unsigned char red = pixelVal.val[2];
        unsigned char green = pixelVal.val[1];
        unsigned char blue = pixelVal.val[0];

        outputRGBValues(red, green, blue);
        if (red == REDCOLOUR && green == GREENCOLOUR && blue == BLUECOLOUR)
        {
            count++;
        }
    }
}
cvReleaseImage(&img);

Also, note that usually OpenCV arrange data in BGR format, which can verified at IplImage::channelSeq. Do not confused by IplImage::colorModel, which tells the color model.

I'll go with Paul R suggestion that you should find a reference and understand the library before attempting to use them.

Answer 2

I've never used OpenCV, but I would imagine that cvCreateImage doesn't initialise the contents of the image, and certainly not to anything meaningful.

Answer 3

Try this:

IplImage* img = cvLoadImage("c:\\test.png");

for (int i = 0; i < img->height; i++)
{
    for (int j = 0; j < img->width; j += img->nChannels)
    {
        unsigned char red = img->imageData[i * img->widthStep + j + 2];
        unsigned char green = img->imageData[i * img->widthStep + j + 1];
        unsigned char blue = img->imageData[i * img->widthStep + j];

        outputRGBValues(red, green, blue);
        if (red == REDCOLOUR && green == GREENCOLOUR && blue == BLUECOLOUR)
        {
            count++;
        }
    }
}
cvReleaseImage(&img);

Answer 4

Quite some issues here:

• You need to fill the image with meaningful data
• Depending on how the image was constructed, you may have swapped red and blue
• You should use unsigned char red
• You declare a vector for each pixel and don't use it