5
$\begingroup$

I am attempting to analyze images using the third order image moment (aka skewness), but I am having trouble figuring out how to do so. As it turns out, there are four different types of third order image moments:

$$ \begin{array}{rcl} \mu_{21} &=& M_{21} + 2\bar{x}M_{11} - \bar{y}M_{20} + 2\bar{x}^2M_{01}\\ \mu_{12} &=& M_{12} + 2\bar{y}M_{11} - \bar{x}M_{02} + 2\bar{y}^2M_{10}\\ \mu_{30} &=& M_{30} - 3\bar{x}M_{20} + 2\bar{x}^2M_{10}\\ \mu_{03} &=& M_{03} - 3\bar{y}M_{02} + 2\bar{y}^2M_{01} \end{array} $$

$$ \begin{array}{rcl} M_{p,q} &=& \mbox{Raw moment of order}\ (p+q)\\ \mu_{p,q} &=& \mbox{Central moment of order}\ (p+q)\\ \end{array} $$

Formulas pulled from Wikipedia - Image Moments.

It's been very difficult to find out what the differences are between these 4 (if there are any) on the internet. If there's anyone here that could explain in layman's terms how they differ, I would greatly appreciate it.

This question is most likely asking something that's more fundamental to the calculation of moments themselves than specifically with regard to image moments. However, I have seen several questions about them here, so I figured it would be a good place to start.

Improve this question
$\endgroup$
2
  • $\begingroup$ It would be good to get the definitions of $M_{kl}$ in your equations! $\endgroup$
    – Peter K. ♦
    Commented Aug 13, 2015 at 12:17
  • $\begingroup$ @PeterK. Appropriate edits made. $\endgroup$
    – cadams
    Commented Aug 13, 2015 at 18:53

1 Answer 1

Reset to default
2
$\begingroup$

It it based on the definition of Central Moments on 2D grid given in the Wikipedia Page.

$$ {\mu}_{pq} = \int\limits_{-\infty}^{\infty} \int\limits_{-\infty}^{\infty} (x - \bar{x})^p(y - \bar{y})^q f(x,y) \, dx \, dy $$

Since we're dealing with 2 random variables, a multiplication which the powers sum to 3 can be achieved in 4 different methods.

It's up to you to decide how to calculate the third moment.

Usually I'd use the Rotation Invariant Moments.
You can find the definition on the same page.
Enforcing the Rotation (And Scale) Invariance yields single definition.

Improve this answer
$\endgroup$
1
  • $\begingroup$ Earlier today, I became familiarized with the rotation invariant moments, realizing that they were what I was needing! Would you happen to know which of the 7 would be able to provide the most effective separation between images containing a small line of high pixel intensity and images with a small curve of high pixel intensity? $\endgroup$
    – cadams
    Commented Aug 14, 2015 at 13:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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