• Advertisement
Sign in to follow this  

Question with gaussian curve

This topic is 3850 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

Hello. I'm using a standard Gaussian curve in order to compute a kernel for my Bloom Filter. The problem is that with the standard Gaussian Curve, the maximal value of the curve vary according to the "Standard deviation" parameter. ( look at : http://en.wikipedia.org/wiki/Image:Normal_distribution_pdf.png to see what i'm talking about ) And because of that, it's hard to correctly define the good behavior for the Bloom filter. How can i get a gaussian curve that still have the same "maximal value" ? Is there a scalar factor by which i can multiply the result of the gauussian equation in order to "re normalize" it ?? Thanks. Clément Vidal

Share this post


Link to post
Share on other sites
Advertisement
Usually, you want no "intensity gain" after the blur (effectively, conservation of energy). This means that your weights must sum to 1, in which case you cannot select the "maximal value" for the center point. You could sample the Gaussian distribution and divide by some factor so that the center point has "maximal value" that you specify, but then your weights most likely will not sum to 1. The result is either an intensiy gain or an intensity loss.

Share this post


Link to post
Share on other sites
You're using a discrete filter, right? So your filter is represented by an array of discrete values? To normalize it, just add up all the values, then divide them all by that sum. The sum of all the values in the array is now 1.

Share this post


Link to post
Share on other sites
Thanks for your help.

I have tryed both solution and, normalizing it using the "second" solution seams to be the better solution :)

Thank again

Share this post


Link to post
Share on other sites
Sign in to follow this  

  • Advertisement