4
$\begingroup$

I am trying to understand granulometry. I understand that it deals with determining the size distribution of particles in an image and uses the principles of opening and closing.

I have a smoothed image with marbles in different sizes and call it A and a structuring element B, which is a disk with variable radii. I get that when opening the image with the structuring element B, smaller marbles disappear and by increasing the radius of B, bigger marbles start to disappear.

What I do not understand is what will happen if I use closing instead of opening. Do the bigger marbles disappear with a smaller radius B? So, does the exact opposite happen with closing?

$\endgroup$
1
$\begingroup$

When you use the opening, the overall operation is called a granulometry. However, when you use a closing, it's called an anti-granulometry. It's the dual operation. Combined together, they form the Pattern Spectrum.

In your case, if you apply a closing, you will certainly measure the structures between the marbles.

However, it is definitely not recommended to apply a smoothing first, because it will smooth the edges and small particles. Consequently, the particles detection will be less accurate.

$\endgroup$
  • $\begingroup$ It depends on the image quality, of course. If the image is fine, and there isn't residual noise present, then no preprocessing is necessary. Otherwise, a low pass filter with window size much smaller than your smallest particle is recommended to smooth out unnecessary details and noise in the image. $\endgroup$ – goldrik Dec 21 '17 at 10:31
1
$\begingroup$

You can think of closing as the dual operator to opening; while opening the image using structuring elemnts $B$ at different sizes gives you values for the Pattern Spectrum $P(k)$ at $k >= 0$, closing will give you values at $k < 0$. These give you an idea of the sizes of "holes" present in the image, rather than the actual objects.

Closing the image is also equivalent to opening the image complement.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.