Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
Drexel University or The Math Forum.


Cameron
Posts:
4
Registered:
3/4/11


Re: anybody got a complex 2D median filter script they can share?
Posted:
Mar 4, 2011 11:42 AM


"Matt J" wrote in message <ikr2rg$345$1@fred.mathworks.com>... > "Cameron " <cameron.matthews@yahoo.com> wrote in message <ikr234$d89$1@fred.mathworks.com>... > > > > > > Thats essentially what I was doing before. I think it violates some part of signal processing to do it that way though and is just filtering out a bunch of stuff I don't want serendipitously. I am trying to preserve a phase characteristic of the complex imaginary data. there are a few papers on IEEE about the application for complex weighted median filters, but frankly I don't think I have time to digest the math and get a script written. > ============= > > So your 2D image is frequency domain data? Maybe you just want to filter the amplitude response and keep the phase, e.g., > > X=medfilt2(abs(X)).*angle(X) > > In any case, I doubt you're going to find anyone on here (quickly) who is familiar with the few obscure IEEE papers that you've alluded to. You'd best start describing your problem in enough detail for people to make suggestions. ================================= Fair enough... I am looking at two views of the same area. my data is coherently sampled so I have phase data. because I am looking at the area from different vantage points I have a different grazing angle and there is speckle associated with the phase offset.
I don't want to discard the phase information entirely so I don't want to look at the magnitude. I found for the application a median filter works very well, but I want analyze the phase data with it and the standard medfilt2 algorithm only works on real numbers.
since phase wraps around, I don't think a conventional median filter will necessarily provide any meaningful data.
SO  I guess an addendum to this request for help would be:
if I median filter the amplitude,then separately median filter the phase or angle and then resolve the two datasets to a complex number, is the math in fact nonvalid?



