[futurebasic] Re: x-fb FFT and/or FIR

Message: < previous - next > : Reply : Subscribe : Cleanse
Home   : December 2001 : Group Archive : Group : All Groups

From: Herbie Gluender <H.Gluender@...>
Date: Sat, 15 Dec 2001 13:08:47 +0100
Robert,

your response to my remark may be convincing, at least for you.

One of the things that don't convince me is:
If you are the expert who knows about Inverse filtering (deblurring), Wiener
optimum filtering, the combination of both, and the like, why then don't you
do it. The correct use of these techniques is much more difficult than writing
a 2D FFT routine which is only about one page of code.

I did (digital) image processing (including all the techniques you have
metioned) 25 years ago when this field was a scientific enterprise. Today we
have a lot of tools that seem to make life much easier but I guess the problem
of understanding the approaches and their applicability hasn't changed that much.


Herbie wrote:

>This may be true in very special cases where you need extended kernels. But
>this doesn't occur very often, except perhaps in very fine arts. :-) The
>Fourier Transformation is of advantage, if you like to calculate the
>autocorrelation function of an image. But why should you want to...

Robert wrote:

Wiener filtering for deconvolution and image deblurring.

If I want to be doing it, I need to be doing it. Knowledge is power :)
Would be neat to paint an image in FFT space, then inverse transform it and
see what you get back out.

Herbie: Do you know how to paint complex numbers?

Kernels in terms of high-pass are usually what are thought of as sharpen
filters...the high pass I am referring to is one similar to Photoshop's
High Pass filter which I understand to be an FFT based item. You can filter
the fringes of FFT images using radial filters easily, works wonders.

Herbie: How do you know that this Photoshop filter is realized in the Fourier domain?

Kernel code I have, a 2D FFT with inverse for images would be handy too.
Beyond my concentration level to comprehend right now unfortunately.

Herbie: I think there is much left to learn...

See "The Computer Image" by Alan Watt, if you need more reasons why I want
to do this.  Lot of examples there of nice applications of FFT and image
processing.


I wish you much progress in your efforts.

Herbie
*************************************************
H.Glu@...