如何获得缩放图像中点的原始坐标？

Question

I have this school assignment which asks me to be able zoom in and out on a region of interest using opencv and C++. The size of the region of interest is initialized at 80*80 and it augments (left click) and diminishes (right click) by 10px vertically and horizontally on click. I think I have managed to do that as so:

// callback associé à la souris sur la fenêtre "TP1"
// clic droit -> zoom arrière
// clic gauche -> zoom avant
// (x, y) indique la position de la souris sur la fenêtre
// event est un évenement de la souris (clic, souris déplacé, etc.)
// flags indique si un bouton de la souris ou du clavier est enfoncé
void cb_on_mouse(int event, int x, int y, int flags, void*)
{
    Mat tmp;    // temporary image
    bool update = false;

    // copying src to tmp
    src.copyTo(tmp);

    // copying src to dst
    src.copyTo(dst);

    // when mouse moves
    if (event == EVENT_MOUSEMOVE)
    {
        // showing x and y coordinates of the mouse on the main image
        putText(dst, "x: " + to_string(x), Point(dst.rows + 4, dst.cols - 25), CV_FONT_HERSHEY_COMPLEX, 5, (0,0,0), 8, true);
        putText(dst, "y: " + to_string(y), Point(dst.rows + 4, dst.cols - 5), CV_FONT_HERSHEY_COMPLEX, 5, (0,0,0), 8, true);
        imshow("TP1", dst);
    }

    // when a click happens
    else
    {
        // if ctrl + mouse_btn then there's no zoom
        if (flags != EVENT_FLAG_CTRLKEY + EVENT_FLAG_RBUTTON && flags != EVENT_FLAG_CTRLKEY + EVENT_FLAG_LBUTTON)
        {
            // left click zooms in
            if (event == EVENT_LBUTTONDOWN)
            {
                // zoom in limit set at 2 for both height and width
                if (roi.height > 20 && roi.width > 20)
                {
                    roi.height -= pas_yzoom;
                    roi.width -= pas_xzoom;
                }
            }

            // right click zooms out
            else if (event == EVENT_RBUTTONDOWN)
            {
                // zoom out limit is set of original image size (or close)
                if (roi.height < tmp.rows && roi.width < tmp.cols)
                {
                    roi.height += pas_yzoom;
                    roi.width += pas_xzoom;
                }
            }

            // calculating x
            // making sure it doesn't get out of bounds
            if (x - (int)(roi.width / 2) <= 0)
                roi.x = 0;

            else if (x + (int)(roi.width / 2) >= tmp.cols)
                roi.x = tmp.cols - roi.width;

            else
                roi.x = x - (int)(roi.width / 2);

            // calculating y
            // making sure it doesn't go out of bounds
            if (y - (int)(roi.height / 2) <= 0)
                roi.y = 0;

            else if (y + (int)(roi.height / 2) >= tmp.rows)
                roi.y = tmp.rows - roi.height;

            else
                roi.y = y - (int)(roi.height / 2);

            update = true;  
        }

        // only updating the image if an update took place
        if (update)
        {
            zoomed_img = tmp(roi);  // creating tmp image
            imshow("Zoom", zoomed_img); // displaying tmp
        }
    }
}

Now I need to gather information about a pixel clicked within the zoom window, but I do not know how to calculate its coordinates on the original image. I really don't know how to deal with scale and all that. I tried this but it isn't working:

Mat tmp_zoom;   // temporary image for zoom
Mat tmp_src;    // temporary image for source
float base_x;   // x on original image
float base_y;   // y on original image
float scale;

zoomed_img.copyTo(tmp_zoom);
src.copyTo(tmp_src);
scale = (float)(tmp_zoom.cols * tmp_zoom.rows) / (float)(roi.width * roi.height);

if (event == EVENT_LBUTTONDOWN)
{
    base_x = ((float)x / scale);
    base_y = ((float)y / scale);

    cout << "Click on zoom: " << "\n";
    cout << "base_x: " << base_x << "\n";
    cout << "base_y: " << base_y << "\n";
    circle(tmp_src, Point(base_x, base_y), 5, (0, 0, 0), 1, 8, 0);
    imshow("TP1", tmp_src);
}

Here's what the program is supposed to do: image_program

I hope that was clear as problem description, thank you in advance for helping.

Answer 1

sorry I didnt review your code.

If you zoomed in a roi you can remember where this roi is placed (position and size).

For example:

roi.x = 300;
roi.y = 200;
roi.width = 20;
roi.height = 20;

Now if you have a resized image holding only this 20x20 region but resized to, let's say 100x100 pixels image, and you display it, you want to click somewhere and compute the corresponding original position?

In that zoomed image you know the click cordinate, like

click.x == 10;
click.y == 80;

Now you can compute the relative image coordinate within the zoomed image like

cv::Point2f relative;
relative.x = (float)click.x/(float)zoomedWidth;
relative.y = (float)click.y/(float)zoomedHeight;

Now in your original image your click pos should be

cv::Point2f original;
original.x = roi.x + relative.x*roi.width;
original.y = roi.y + relative.y*roi.height;

hope this helps