The fastest way to Recognize a shape from background in Open CV, C++

Question

I Have an image as background, and an image from the background with the shape in it. I want to recognize the shape as fast as possible. One of the things I tried was:

Mat img, img_background;
.
.
.
img = img - img_background;

But in some images, it didn't work. There was another way to compare all of the pixels in both images and it worked, But it was pretty slow. I want to know is there a way to recognize the shape as fast as possible? Thanks for your help.

---

And I have a question. Why the code I wrote(img = img - img_background) didn't work?

---

I could fix it with the code below:

   cv::absdiff(img, img_bk, img);
   Mat m = Mat::zeros(img.size(), img.type());
   m.setTo(Scalar::all(255));
   img = m - img;

I guess this code works. Thank you for your helps.