"Olli Niemitalo" <o@iki.fi> wrote in message
news:Pine.GSO.4.58.0310200558460.289@paju.oulu.fi...
> On Sun, 19 Oct 2003, walala wrote:
>
> > [...] I am doing "floating-point" search and trial & error ... so the
> > search space is huge. I just try to search for a few days...
>
> If you want to make this time much shorter, try evolutionary algorithms.
> We have written a small c++ library for that:
>
>   http://www.iki.fi/o/opti/ok
>
> -olli

Olli,

Thanks  a lot for the pointer... I have looked at the program but have a
problem:

Basically I am doing everything in Matlab, since I need the
easy-manipulation of image data, and the easy-scripting, and the filtering
operation, etc. And your program is in C++... I tried to compile Matlab
script file into standalone executable and C++ source codes and then combine
with yours... Before I can see any results, it just give me a lot of error
during compilation. Also the C++ source code generated by Matlab is quite
weird and not understandable. I wonder how to combine with yours. I suggest
you to make a Matlab version or at least make a Matlab interface for your
optimization code. Otherwise people may need go through a lot of trouble
before they can try on your optimization code.

Thanks a lot for your work!

-Walalal

"walala" <mizhael@yahoo.com> wrote in message news:<bmvju8$dgl$1@mozo.cc.purdue.edu>...
> >
> > Wiener filters come to mind... again, there are plenty filters and tricks
> > available for image processing. Whether you can do stuff completely
> > automatic or need a human operator in the loop, depends on the
>  application.
> >
> > Rune
> 
> Rune,
> 
> You said it... after doing searching a lot, I am tired of trial and error.
> This cannot let us write a report. I will try to apply a Wiener filter...
> just headache about how to get the Rxx of the input image...which is the
> covariance matrix of the input? Can you give me information on this?

Sorry, can't help. I am not even sure Wiener filters work for image 
processing, but somehow I think I have seen the term used with images...
perhaps you should ask at an image processing newsgroup?

As for writing reports... I have the extremely controversial view that 
a report should reflect actual progress, not some imaginary path towards 
the desired goals. Some times the solving of a problem is just too 
difficult or require more resources than available in a project, and I 
believe the report should reflect that. That way there is a chance that 
planning a next project over the same theme has a chance to succeed.
If used properly, such progress reports would actually be a great tool 
to show which managers actually know their job and which would be better 
off somewhere else. 

But beware! These counterproductive "non-team player" attitudes of mine 
forced me to resign from one job and got me into a long sick leave from 
the next. You may want to play your cards differently.

Rune

>
> Wiener filters come to mind... again, there are plenty filters and tricks
> available for image processing. Whether you can do stuff completely
> automatic or need a human operator in the loop, depends on the
application.
>
> Rune

Rune,

You said it... after doing searching a lot, I am tired of trial and error.
This cannot let us write a report. I will try to apply a Wiener filter...
just headache about how to get the Rxx of the input image...which is the
covariance matrix of the input? Can you give me information on this?

Thanks,

-wa

On Sun, 19 Oct 2003, walala wrote:

> [...] I am doing "floating-point" search and trial & error ... so the
> search space is huge. I just try to search for a few days...

If you want to make this time much shorter, try evolutionary algorithms.
We have written a small c++ library for that:

  http://www.iki.fi/o/opti/ok

-olli

jim" <"N0sp wrote:
> And most important the compression in jpegs is done in part by quantizing
> the frequency levels of each block.

...which is one of the reasons why a simple low pass filter increases PSNR.
I guess predominantly however, it is because it reduces the high contrast
edges, which is where the error is greatest.  Since these edges tend to be
the worst perceived artifact, low-pass filter helps drastically there.
(Bearing in mind that PSNR is not entirely equivelent to perceived image
quality)

Ben
-- 
I'm not just a number. To many, I'm known as a String...

walala wrote:

> The input images are some jpeg images having block-artifacts. As you know
> that JPEG images have block-artifacts under low bit rate. Someone else in my
> group did some experiments to pre-filter the images to try to get rid of the
> artifacts. But that also introduces other artifacts. So now it is a combined
> artifact... and to be frankly, all processing techniques looked the same to
> me on images, because they are nearly same, but in terms PSRN, they have big
> difference... so I trial and error and found one with both good PSNR and
> looking(but still ugly though)...

Okay, now you are providing the type of information needed. So the task is
to remove the blocking artifacts from jpegs. Here are some things you may
want to consider:
	Jpegs have different levels of compression resulting in different levels
of blocking artifacts.
	The compression techniques used in jpeg involve analysis of the frequency
content of the image and the amount of blocking artifacts is somewhat
related to the frequency content of the original input. Jpeg is designed
for natural images like photographs. Artificial images like line drawings
tend to have a lot more objectionable artifacts.
	And most important the compression in jpegs is done in part by quantizing
the frequency levels of each block. This is done by formulae (derived by
experiment) to retain the part of the signal that is the most perceptually
important. This explains why you can't see the difference even though your
calculations show a big difference in the amount of noise removed. 

	Another thing is you are not limited to 3 x 3 filter. A common filter
I've seen applied to this problem is [-1 0 5 8 5 0 -1]/16. This filter can
be applied very efficiently with adds and shifts by iterating through the
columns and rows of the image, resulting in what is effectively a 7X7
filter. I don't know if speed is an issue you are interested in.

-jim
	

> 
> But I cannot do trial-and-error search for every input, right? Because after
> this "training" stage, I don't have source images available any more, right?
> My fitler should work adaptively itself...


-----= Posted via Newsfeeds.Com, Uncensored Usenet News =-----
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
-----==  Over 100,000 Newsgroups - 19 Different Servers! =-----

walala wrote:
> "Ben Pope" <spam@hotmail.com> wrote in message
> news:bmuqba$qdnh9$1@ID-191149.news.uni-berlin.de...
>> walala wrote:
>>> Are there any guidelines on how to design good filters that applicable
>>> to most(if not all) images? Discussion should be in two cases: a) with
>>> source image and can get PSNR to compare results; b) with no source
>>> image and only have the reconstructed images, so no PSNR can be
>>> computed, how to design a "good" filter fit into this case?
>>
>>
>> You need to first work out your metric... you say you use PSNR, but
>> without a source image... so how can you possibly measure PSNR?
>>
>
> Bob, thanks for your posting,

Who is Bob?

> That's the problem. The filter is supposed to work on all images. In this
> "design" stage, I have a few source images to test PSNR... later on, the
> filter is supposed to adapt to input and do filtering blindly... at that
> time, no PSNR can be measured.

Right.

> So what I want to know is exactly how to design a filter that can
> adaptively work on images to achieve satisfactory PSNR(although may not
> be the best, but al least works for most cases) after the initial
> training stage.


Well if you're trying to clear up JPEG artifacts, then the best you can do
is some kind of low pass filter.

Wavelets can be used to great success as well.  But thats whole new story
:-)

PSNR is not a great measure of perceived image quality anyway.  We are more
sensitive to brightness than colour, by a factor of about 4, yet PSNR does
not model that.  Additionally the edges that you see with JPEG artifacts is
also not well described with PSNR, human vision is designed to pick out
edges, we're very good at it, so although the error might be quite small in
terms of PSNR, it looks ugly to us.

Go back to my original reply... start defining your problem, are you trying
to remove JPEG artifacts from images?  Then the biggest issue I have with
JPEG images are the discontinuities... so low pass filter it, which is what
you have been doing.

In terms of the specific best filter... you're fighting a losing battle
without a huge training set.  If you're images are primary photograph based,
then just use Gaussian.  Otherwise do a search for papers that have
researched JPEG artifact removal - I suspect there are a considerable number
of them out there with a wide ranging number of techniques.

You can remove many types of noise from an image quite effectively with a
simple blur (low-pass filter, guassian being very popular, along with
unsharp mask), to increase PSNR by a good 10dB or so.

Ben
-- 
I'm not just a number. To many, I'm known as a String...

"Ben Pope" <spam@hotmail.com> wrote in message
news:bmuqba$qdnh9$1@ID-191149.news.uni-berlin.de...
> walala wrote:
> > Are there any guidelines on how to design good filters that applicable
to
> > most(if not all) images? Discussion should be in two cases: a) with
source
> > image and can get PSNR to compare results; b) with no source image and
> > only have the reconstructed images, so no PSNR can be computed, how to
> > design a "good" filter fit into this case?
>
>
> You need to first work out your metric... you say you use PSNR, but
without
> a source image... so how can you possibly measure PSNR?
>

Bob, thanks for your posting,

That's the problem. The filter is supposed to work on all images. In this
"design" stage, I have a few source images to test PSNR... later on, the
filter is supposed to adapt to input and do filtering blindly... at that
time, no PSNR can be measured.

So what I want to know is exactly how to design a filter that can adaptively
work on images to achieve satisfactory PSNR(although may not be the best,
but al least works for most cases) after the initial training stage.

"walala" <mizhael@yahoo.com> wrote in message news:<bmuae2$pbe$1@mozo.cc.purdue.edu>...
> You mean all experts do filter design by ad hoc search and trial and error?

Eh... I can only answer for myself and the people I have worked with...
but yes, quite a lot is trial and error, some times completely ad hoc, 
other times based on experience or educated guesses. When presented with 
a problem, there is usually more than one way of doing things. In my 
opinion, one part of the job is to know as many such tools and tricks and 
methods as possible, the other part is to evaluate which methods works 
how well and when. If you look in the only book I know of that teaches 
data processing, 

  Yilmaz: Seismic Data Processing 
    SEG, 1987

you will find a lot of more or less standard flow charts for doing 
various standard processing tasks. But in a seismics processing lab the 
analysts spend days and weeks trying to fine-tune the processing 
parameters. Some of these guys are truly artists. 

Actually, someone I worked with did a test where several different 
processing teams got the same data set and were assigned the same 
processing tasks. When the results came back, you almost had to 
know in advance that these images were supposed to show the same 
geologigal features. They almost looked like data from different 
oil fields. Data processing is a very subjective dicipline.

> What I need is pointers/guidelines on systematically designing a filter
> based on the input. Since although the source is unknown, yet the
> reconstructed input still has some statistical property. If some
> mathematical analytical model can be made based on the parameters extracted
> from the input, then a filter can be adapted to those parameters for that
> particular input.

Hmmm... this depends on the application. Be aware that you, by choosing 
this analytical model, puts constraints on the resulting images. If you 
somehow gets an image that does not fit that model, you may be in trouble.
If, for instance, you work with medical images and search for soft tissue 
injuries, a processing algorithm could use known information about e.g. 
the skeleton, and use deviations from the "expected" image to assess 
tissue damages. But what if there is severe damages to the bones as well?
If the processing algorithm assumes an undamaged skeleton, the automatic 
processing routine will get lost.
 
But then, if your images are sufficiently regular, like what could be 
expected in certain quality control tasks on a production line, such 
variations need not be a problem at all.

> Since it turns out already that the "best" filter for a source is "near
> Gaussian" and it is small, 3x3, I believe the "best" filter for anohter
> source would also be "near Gaussian" but with adjusted numbers.... So is
> there any pointers/guidelines on mathmatical modeling for derivation of such
> kind of filter?

Wiener filters come to mind... again, there are plenty filters and tricks 
available for image processing. Whether you can do stuff completely 
automatic or need a human operator in the loop, depends on the application.

Rune

> Ok, so the filter must be symetrical - this is not exactly the same as
> gaussian. What you haven't yet explained is how you are determining that
> one filter works better than another.

The filter is supposed to work on input images and enhance them. I have
several images, but not all. So I tried on these a few images on my hand and
"Gaussian-like"(as the following one)  filter has great PSNR comparing with
"averaging" or other filters(e.g. 39dB vs. 10dB)... since I have source
images so I can computer PSNR and compare, right?

>
> > I finally got:
> >
> >    -0.0496    0.1408   -0.0496
> >     0.1408    0.6355    0.1408
> >    -0.0496    0.1408   -0.0496
> >
> Its low pass. Its not really gaussian with the negative values. Somewhere
> half way beteen an unsharp mask and gaussian.
>

Ok, I see.

> >
> > I believe my filter should be adaptive, hopefully the numbers in this
filter
> > can change according some underlying properties in the input source.
How to
> > do that?
> >
> Well, it sounds like you are already doing it by brute force. Maybe if you
> explain what the input source is and how you determine that the filter
> "works", then some one might suggest ways to optimize your search.

The input images are some jpeg images having block-artifacts. As you know
that JPEG images have block-artifacts under low bit rate. Someone else in my
group did some experiments to pre-filter the images to try to get rid of the
artifacts. But that also introduces other artifacts. So now it is a combined
artifact... and to be frankly, all processing techniques looked the same to
me on images, because they are nearly same, but in terms PSRN, they have big
difference... so I trial and error and found one with both good PSNR and
looking(but still ugly though)...

But I cannot do trial-and-error search for every input, right? Because after
this "training" stage, I don't have source images available any more, right?
My fitler should work adaptively itself...