CODEWITHSUNDEEP
c++opencvimage-processing
Dan Prince
Dan Prince

Reputation: 29989

Wrong RGB values when modifying image with OpenCV

Fairly new to C++ and I've been working with the OpenCV library. I'm attempting to create a dimming function for an image viewer. This is the main body of void dim(int val).

I'm looping over every pixel, getting a reference to the RGB values of the pixel within a Vec3b then decreasing them by some change factor, calculated from val / 255 (where 0 <= val <= 255. But for some reason the pix[0] etc are all being set to 0, meaning that for any value of val the image turns black.

May be doing something slightly stupid with floating point here, or misunderstanding some aspect of OpenCV; but I can't see what it is.

for(int x = 0; x < rows; x++) {
    for(int y = 0; y < cols; y++) {
        Vec3b &pix = dst.at<Vec3b>(x,y);

        float change = val / 255.0f;
        pix[0] = (uchar)(pix[0] * change);
        pix[1] = (uchar)(pix[1] * change);
        pix[2] = (uchar)(pix[2] * change);
    }
}

Thanks in advance!

Upvotes: 0

Views: 657

Answers (2)

WangYudong
WangYudong

Reputation: 4413

OpenCV has a doc about changing the contrast and brightness of an image. Generally speaking, it applys alpha * pixel_value + beta to every pixel. Since you only need to adjust the brightness of images, alpha can be set 1.0. Here is the modified dim() function.

// negative beta dims image
void dim(Mat& src, Mat& dst, int beta) {
    for(int y = 0; y < src.rows; y++) {
        for(int x = 0; x < src.cols; x++) {
            for(int channel = 0; channel < 3; channel++) {
                dst.at<Vec3b>(y, x)[channel] = saturate_cast<uchar>(src.at<Vec3b>(y, x)[channel] + beta);
            }
        }
    }
}

Read what saturate_cast does if interested.

To test the function above:

int main( int argc, char** argv )
{
    int beta;
    Mat src = imread(...);
    Mat dst = Mat::zeros(src.size(), src.type());
    namedWindow("src", CV_WINDOW_AUTOSIZE);
    imshow("src", src);
    dim(src, dst, -200);
    namedWindow("dst", CV_WINDOW_AUTOSIZE);
    imshow("dst", dst);

    waitKey(0);
    return 0;
}

Upvotes: 1

Andrey Smorodov
Andrey Smorodov

Reputation: 10852

If you multiply/divide integer value by real value less than 1 you'll get 0 because your scaling coefficient will be truncated to integer. To correct your problem you should convert pix[] to float or double when scale it and convert result back to char.

Upvotes: 0

Related Questions