Adaptive vs Non Adaptive Filters

I am struggling to understand the difference between adaptive and non adaptive filters in general. Also, in terms of the Wiener filter, I have seen it referred to as both an adaptive and a non adaptive filter. So what makes a filter adaptive or non adaptive? Thanks

Answer 1

[Shashank Prasanna]

I think wiki sums it pretty well:

http://en.wikipedia.org/wiki/Adaptive_filter

[B]

I guess my question applies more to the Wiener filter. Wiki says that it is non adaptive ( http://en.wikipedia.org/wiki/Wiener_filter) and MatLab says that it is adaptive ( http://www.mathworks.com/help/images/ref/wiener2.html)

[Image Analyst]

Not exactly. MATLAB says that the wiener2 filter (not the wiener filter) "filters the image I using pixelwise adaptive Wiener filtering". So wiener2 uses wiener but changes it (adapts it) as it slides along the image.

[B]

Ok, the way I understand it is that the Wiener filter is an inverse filter that takes a corrupted image and attempts to deconvolve the uncorrupted image and a degradation function (PSF) in the presence of additive noise. I believe it does this my minimizing the mean square error. The solution for the estimation of the uncorrupted image that results in the minimum MSE is shown in the first equation here http://www.owlnet.rice.edu/~elec539/Projects99/BACH/proj2/wiener.html If I wanted to make the filter adaptive, where in that equation do I take into account the local variance? Is it in the power spectra?

Answer 2

[Youssef Khmou]

hi, i think that :

The adaptive filter is usually linked to dynamic systems , the time is involved in the equations, so the filter performs the filtering actions at time t (t+1) based on posterior state t-1 (t), this way the filter adapts itself to new conditions it meets meaning that if the signal to be filtered has harsh fluctuations the filter operates based on those fluctuations,

Non adaptive filter, is a simple protocol ( algorithm) that can be applied to the signal without computing some statistics,

The problem of adaptive or not is based on Frequency Response ( Finite or not) and the intuition is that the adaptive filter performs well better that the non adaptive one .

The topic is open for discussion .