[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Re: [APML] some comments



>From: "Tom & Lou Krajci" <krajcit@3lefties.com>
>Reply-To: astro-photo@seds.org
>To: <astro-photo@seds.org>
>Subject: Re: [APML] some comments
>Date: Sun, 23 Sep 2001 10:48:01 -0600
>
>From: <Tonyhallas@aol.com>
>
> >... I can explain what I doing WITHOUT trying to define it...
>
>This may help define it in more mathematical terms:
>http://www.astropix.com/HTML/J_DIGIT/COMPEXP1.HTM   ...yes, the effect is
>most noticeable on the sky background and other faint features.

from http://www.astropix.com/HTML/J_DIGIT/COMPEXP1.HTM :
>Theory
>The technique works because the signal in each image is constant while the 
>noise is random. When >the multiple images are composited together the 
>noise sort of cancels itself out while the signal >remains the same. The 
>more images that are combined, the greater the signal to noise increase.


I have a question for Tony, Tom, Jerry Lodriguss (the author of the above 
page on "composite"), & others on APML..

In the above statement By Jerry Lodriguss, what do you consider "signal" & 
what do you consider "noise"?

Once you tell me that, then I can determine what you're "thinking" in terms 
of "signal to noise ratio", & I can help you *formulate* a valid 
description.  This would be the proper step, in order to get amateurs on the 
right track, in terms of *understanding* the "noise" issue in film-based 
astro images.  Once we get to that point, we can start developing novel 
algorithms (as per Dr. Joe Pedit), to attack the problem.

>From: <Tonyhallas@aol.com>
>
> >... I can explain what I doing WITHOUT trying to define it...

Oh boy.. IF you can't DEFINE it, why are you making statements about it??  
Come on!  Wasn't it Jerry Lodriguss who said on APML, "Oh, I love 
definitions!!".  An image processing scientist (a qualified one, that is) or 
a mathematician should have been consulted, to proofread your statements or 
article.  (The latter is kinda of a basic thing anybody should do, to check 
your work.)  That's whats missing in www.astropix.com, this thing is filled 
with so many errors, it's UNVBELIEVABLE.  If Tom Krajci/Air Force 
Intelligence Officer got fooled by it..I'm SCARED TO DEATH.  (I hope to God, 
you ain't using it for Air Force reconnaissance image analysis.  Please tell 
me you AREN'T) Now, I can understand how American Intelligence (CIA, FBI, 
NSA) failed in preventing the disaster of last week.  (see 60 Minutes, "How 
Intelligent Were We?".  Turns out the CIA and FBI really "dropped the 
ball").  I've always suspected the entire APML (many technically trained 
people, incl some PhDs) got fooled by Jerry Lodriguss' "pretty pictures" and 
neatly laid out "reference".  Bottomline: it has no peer-review by optical 
engineers, image processing scientists..it's a meaningless datapoint.

The French amateur Thierry Legault (member of the school of High-Resolution 
Astrophotography, as per Jean Dragesco, who is world-renowned) uses 
*hundreds* of samples (sub-second integration times, to "freeze" the 
seeing), throwing away bad samples, followed by statistically-based filters 
(*median-filter* image combining, for example).  His "Hubble-like" images of 
Saturn, Jupiter are a testament to the validity of his approach.  Basically, 
he gets *quality* images which practically eliminate "seeing" (atmospheric 
effects), followed by REAL noise-removal using statistically-based filters.  
The latter involves NON-AVERAGING operations, which is the key.  The PW 
"blend" function is basically arithmetic averaging, which is not an optimal 
way of "noise removal".  IT's based on the underlying-theory of sqrt(N), N=# 
of  images, S/N ratio increase, which itself is a mis-quoted & mis-used 
result in the image-processing literature (which has unfortunately, been 
further twisted by amateurs like Jerry Lodriguss.  The weblink quoted above 
is FILLED with errors, contradictions, misinformation.  It would take a 
detailed essay, to address each mistake.  I'm afraid I don't have the time).

Simple Fact: _Averaging_ is a *dangerous* operation in math, science, or any 
field..

Question:
Would any of you send your kids to an "average" public high school, & 
subject them to the "averaging effects" of mediocre teachers and mediocre 
(or below average) students??

Answer:
Of course not!  You buy homes in good real estate markets, with good 
schools.  (You may even pay big bucks to send them to private schools).  You 
pay good  $$ to get a quality education for your kids..you're aiming at the 
"upper tail" of the Gaussian Curve.  "Averaging", will put you lower on the 
Guassian curve.  (Half of the people are "Below Average", based on the 
Guassian Curve)

So, SAME THING..for combining astrophotographic film images (or CCD, for 
that matter).  Don't use averaging..it's BAD NEWS.  The goal is to 
"penetrate" the statistical "noise barrier", to *measure* the "real signal". 
  The question is how to do it?  Answer: statistical filters.  However, with 
2 to 3 frames..you are REALLY UNDERSAMPLED.  You need MORE samples, as per 
Thierry Legault.  Hundreds of images.  The prospect of taking a 100 hundred 
film (or CCD) images is not a pleasant thought.

To Tony Hallas:

I believe I heard thru someone else, that you believe Jerry was a resource 
and you take his advice.  One word, DON'T.  Your article in Sky&Telescope a 
couple years back reflect his erroneous statements on signal to noise ratio. 
  This is a continuing pattern of paper-trail (traceable back to 
www.astropix.com), which can be used against you.  I'm afraid this could 
ruin your credibility.  Didn't I also hear that Daphne said that her 
impression was that she thought that Jerry just copied other people's 
methods, and took credit for it by publishing it on his website?  (Mr. 
Copycat) She's exactly right, I'm afraid.  That's why there are so many 
mistakes, he's just copying stuff ("knowledge consumer"), not understanding 
("knowledge creation") what he writes.

"Knowledge Consumer" VS "Knowledge Creation"

"I cannot Understand, what I cannot Create"
-- Dr. Richard Feynman, Caltech physicist, Nobel Laureate

APML has always had periodic flareups, of poor signal to noise ratio.  Brad 
Wallis/JPL, Jason Ware (engineer), Wil Milan (engineer) have all left 
because of this.  Some newer members with technical backgrounds, have also 
brought the issue up, of the "chat-room" atmosphere.  As per, "this is my 
image, do you like it?  <followed by 10 responses like.."I love it Dude!">  
I like his image, but I don;t like yours".  This sounds like women's 
gossip!!  This is not consistent with APML charter "To disseminate knowledge 
of Astrophotography".


> >...I am fascinated by the light gathering capabilities of CCD's but can't
>afford a 2K X 3K instrument right now ... so that takes care of that.
>
>Well, to start learning some CCD basics I suppose you could get a less
>expensive, small chip CCD...and use it when springtime skies come into view
>and galaxies are top on the menu.
>
>Tom Krajci
>
>
>--  APML Archives at <http://astro.umsystem.edu/apml/>  ---
>              Unsubscribe at <majordomo@seds.org>


_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp


--  APML Archives at <http://astro.umsystem.edu/apml/>  ---
             Unsubscribe at <majordomo@seds.org>