What is difference between (CountNonZero) and (Moment M00) and (ContourArea) in OpenCV?

Question

If I have a 3x3 binary image and there is a contour in locations(x,y): (0,0), (0,1), (1,0), (1,1)

I get the contour via findContours method.

I want to get this contour's area:

• with CountNonZero: 4
• with contourArea: 1
• with Moment M00: 1

What is the correct answer and what is the difference between them?

This contour is square so the area is 2*2 = 4

So why is ContourArea equal to 1?

I am using EmguCV and this is my code:

VectorOfVectorOfPoint cont = new VectorOfVectorOfPoint();

    Image<Gray, byte> img = new Image<Gray, byte>(3,3);

    img[0, 0] = new Gray(255);
    img[0, 1] = new Gray(255);
    img[1, 0] = new Gray(255);
    img[1, 1] = new Gray(255);
    CvInvoke.FindContours(img, cont, null, Emgu.CV.CvEnum.RetrType.External, Emgu.CV.CvEnum.ChainApproxMethod.ChainApproxSimple);

    Moments m = CvInvoke.Moments(cont[0], true);
    Console.WriteLine(CvInvoke.ContourArea(cont[0]));

    CvInvoke.Imshow("ss", img);
    CvInvoke.WaitKey(0);

Answer 1

I’m not aware of implementation details, but I would suspect that the “contour” is a polygon that goes from pixel center to pixel center around the object. This polygon is smaller than the set of pixels, each edge is moved inwards by half a pixel distance.

This is consistent with the area of a 2x2 pixel block being measured as 1 pixel.

If you want to measure area, don’t use the contour functionality. Use connected component analysis (object labeling) and count the number of pixels in each connected component.

---

OpenCV is not meant for precise quantification, and there are lots of things in it that don’t make sense to me.