grayscale a ROI of a BGR image

Question

I am stuck with a little problem with OpenCV.

I am able to draw a rectangle on my webcam's capture in order to draw a ROI. I would like to know if it is possible to grayscale this part of the frame.

I tried many different ways to do it, but I still can't make it.

Any suggestions?

Answer 1

Here's an example

cv::Mat image, roi_image;
image = cv::imread("example.jpg");  
cv::Rect r = cv::Rect(10,10, 50,50); // 'r' defines the location of ROI
roi_image = image(r);                // select ROI 
cvtColor(roi_image, roi_image, CV_RGB2GRAY);  //modify ROI
imshow("Ouput", image);   // desired output image where ROI is in grayscale
cv::waitKey(10);         // add this line so that graphics can update

Note that roi_image is a matrix that is pointing to the ROI of image. If you modify roi_image it will also modify image.

Answer 2

Here is a crude way to set captured image ROI to grayscale:

cv::VideoCapture capture;
cv::Mat frame, grayFrame, gray3;

if(!capture.open(-1))
{
    cout<<"Capture Not Opened"<<endl; return;
}

int width = capture.get(CV_CAP_PROP_FRAME_WIDTH);
int height = capture.get(CV_CAP_PROP_FRAME_HEIGHT);

cv::Rect roi(20,20,400,400); //The ROI to convert to gray

cv::Mat mask = cv::Mat::zeros(height,width,CV_8U);
for(int i = roi.y; i<roi.y + roi.height - 1; i++)
    for(int j= roi.x; j<roi.x + roi.width - 1; j++)
        mask.at<uchar>(i,j) = 1;

do 
{
    capture>>frame;

    if(frame.empty()) break;

    cv::cvtColor(frame,grayFrame,cv::COLOR_BGR2GRAY);

    cv::cvtColor(grayFrame, gray3, cv::COLOR_GRAY2BGR);

    frame.setTo(cv::Scalar::all(0),mask);
    cv::add(frame,gray3,frame,mask);

    cv::imshow("Image",frame);
    cv::waitKey(10);
} 
while (true);

I couldn't find a simpler way to set image values using masked operation, so the alternate is to set the ROI to zero, and add the masked values to it.

Answer 3

Thanks for your time and patience guys. I'm reading the "Learning openCV" book from O'reilly and all the examples are done with the IplImage class instead of the Mat class. Actually, I don't even know what's the difference.

Anyway, I couldn't use your solution since I do not use any Mat object, here is my solution (it works now) :

while(1)
{
 frame = cvQueryFrame(cam);
 temp = cvCloneImage(frame);
 if(!frame)break;
 cvCopy(frame,temp);
 if(drawing_box || box_drew)
  draw_box(temp,box);

 if(grayScaleOn && box_drew)
 {
  for(int y=box.y;y<box.y+box.height;y++)
  {
    uchar* ptr = (uchar*)(temp->imageData+y*temp->widthStep);
    for(int x=box.x;x<box.x+box.width;x++)
    {
ptr[3*x+0] = ptr[0] * 0.114 + ptr[3*x+1]*0.587 + ptr[3*x+2]*0.299;
ptr[3*x+1] = ptr[0] * 0.114 + ptr[3*x+1]*0.587 + ptr[3*x+2]*0.299;
ptr[3*x+2] = ptr[0] * 0.114 + ptr[3*x+1]*0.587 + ptr[3*x+2]*0.299;
    }
  }
 cvShowImage("Output", temp);