«

>Today Bart van der Wolf commented courteously on the subject
>> That is not correct. With similar techniques as were used
>> to restore Hubble Space Station's initial imagery, it is
>> also possible to restore some of the OOF information.

All Things Mopar <> writes:
>With all due respect to someone I don't know, my first
>reaction to your contradiction of my simple statement is,
>well, horseshit.

Maybe, but you're wrong. Try searching for "deconvolution".

>What did it cost to fix Hubble, several /billion/ dollars?
>And, they did /not/ fix the out-of-focus images /after/ they'd
>been downloaded to earth, those are toast (yes, they tried and
>tried and tried, but never succeeded to any measurable
>degree).

They did succeed in sharpening many images in the several years before
the Hubble optics were fixed. The Hubble was unable to shoot images of
dim stars, one thing it was originally designed for, because the light
was spread over so many pixels. But bright objects could be imaged and
the defocus largely corrected, because they knew the telescope's blur
function so well.

And yes it cost a lot to fix Hubble, but that was a hardware fix. One
camera module was replaced by a very complicated mirror assembly that
corrected the error in the primary mirror. That doesn't mean the software
sharpening of the blurred images cost this much.

>NASA engineers and astronomers colaborated to add software to
>Hubble and some electronics (as I recall, but I'm no Hubble
>expert) to compensate - not correct - the incorrect mirror
>curvature grind to allow /new/ images to be in-focus, and then
>through some pretty sophisticated - read: extremely expensive
>mathematical techniques.

You're just making this up, aren't you?

No, the correction device was hardware only. Once installed, the
telescope delivered sharp images to the various sensors without any
postprocessing needed. As I understand it, the added optics were
arranged to image the primary mirror onto a special correction mirror.
Because of this, all the light that reflected from a particular point on
the primary mirror was re-gathered at a single point on the correction
mirror. That allowed the error in the primary mirror to be corrected by
a complementary correction in the new mirror. All the light from a
single star was focused on a tiny spot a few pixels in size, as
designed.

>And, the space inside Hubble the
>astronaut had to work in was so tight and the chance to
>irreparably damage the telescope was so high that nobody
>really knew in advance if Hubble could or could not be fixed.

Really, what's this got to do with software sharpening?

>> The result will not be perfect, because the truely lost
>> information will generate artifacts during the restoration
>> process. It is also important to have a good model for the
>> de-focus. Some so-called "blind" deconvolution algorithms
>> estimate the blur function, other methods require prior
>> input of the model to be used.

>What kind of techno babble is this?

It makes perfect sense if you'd bother to do any kind of search for
deconvolution methods.

>I'm
>supposed to know in advance exactly how I blew it, so the
>software knows how to begin? If I'm misunderstanding you, I
>apologize but I fail to see how someone can predict the way(s)
>to blow a focus.

It could be as simple as having a bright point source in the image (the
reflection of the sun off a small metal ball, for example) which allows
estimating exactly what the blur function looks like. If you know what
the blur did to the image, you can invert it (with the caveats that Bart
mentioned).

As you point out, all this may not be worthwhile in most cases in
general photography. But you as much as called Bart a liar, while at
the same time mangling much of what you said about Hubble.

Dave

On Tue, 21 Mar 2006 20:59:54 -0600, All Things Mopar wrote:

Today All Things Mopar once again showed his inability to comment
courteously on the subject at hand:

>> That is not correct. With similar techniques as were used
>> to restore Hubble Space Station's initial imagery, it is
>> also possible to restore some of the OOF information.
>
> With all due respect to someone I don't know, my first
> reaction to your contradiction of my simple statement is,
> well, horseshit.

With all due respect, you've once again shown yourself to be a
horse's ass. There are a number of people here that know far more
than you about many areas of photography, and Bart is one of them.
But you wouldn't know that, as you seem to value typing much more
highly than reading and comprehension.


> Before you or someone else decides to take me on about this,
> please keep this in mind: I am a pragmatist, not a
> theorotician. I am also not an elitest. I deal in reality.

Your reality is often the purest fantasy.


> Two sayings come to mind here - "the proof of the pudding is in the
> eating" and "I don't know anything about art, but I know what I like.

You've evidently developed a taste for lip-smackin BS.


> Do whatever floats your boat and I'll do the same. And, you
> have a good evening, hear?! <grin>

And I'll leave you to smile as your boat floats down that well
known river immortalized by the Fugs. <grin and bear it>

Today D-Mac commented courteously on the subject at hand

> All Things Mopar wrote:
>>> What you did here was what I previously said - you
>>> created the /illusion/ of in-focus, and along with it,
>>> some really nasty defects in the background looking like
>>> some weird cross between noise and JPEG artifacts. You
>>> even said in the paragraph above that the "restoration"
>>> process creates those artifacts!
>>>
>
> I'm fed up with contradictions about image manipulation but
> I can't help pointing out to you - whoever you are - that a
> little knowledge is dangerous. You do not comprehend how an
> advanced process can sharpen an out of focus image. If you
> did, you would not be so quick to get up the nose of Bart.
>
> Sharpness is a perceived thing. I resharpen out of focus
> and slightly motion blurred images by recognizing the
> perception and actually blurring a lot of the image even
> more than the amount I correct. The result is an image
> which is perceived to be sharp(er).

No ****?! Thanks for an elitetist, legend-in-their-own-mind,
"expert" putting a dumb-ass old engineer in their place. You
get some sort of jollies debating theoretical horseshit and
taking on people for no good reason? Why not try helping
people in a /practical/ way to fix their problems instead of
making yourself look smart and the other guy look stupid?

> The process is used by NASA although their routine is a
> little different to mine. I convert edges to vector and
> blur the bitmap, re-applying the vector after it has been
> narrowed. This produces the perception of sharpness due to
> the fuzz from the edges being gone and the fill detail
> being smudged.
>
> Although Photoshop has all the tools for doing this, you
> might not get the results I do. Go ahead and try it. Even
> if you stuff it up (as you no doubt will in PS) you will
> learn that it is indeed possible.

No ****?! I didn't know that, either. 'Course, you're probably
bright enough - you know everything about everything - to
recognize sarcasm when you see it...

Look, there is more than simple Gaussian blur, /perceived/
sharpnness and detail here. Take your elitist ideas and re-re-
re-re-review the concepts, early adoption, and later
standardization of COF (Circles Of Confusion) as relates to
the eye's ability to resolve fine detail then come back and
talk about the mathematics behind creating an illusion of
sharpness across /all/ of the photo, including mathematically
correct DOF through COF in front of and behind the point
focused and explain in concrete terms an old fool can
understand just exactly how you take a piece of mush out-of-
focus image and make it look like the camera-photographer
combo did their job right the first time.

--
ATM, aka Jerry

"Whether You Think You CAN Or CAN'T, You're Right." – Henry
Ford

Today Roger N. Clark (change username to rnclark) commented
courteously on the subject at hand

> All Things Mopar wrote:
>
>> Today Bart van der Wolf commented courteously on the
>> subject at hand
>>
>>>"All Things Mopar" <> wrote in message
>>>news:Xns978DA5628F915ReplyID@216.196.97.131.. . SNIP
>>>
>>>>I haven't tried this particular product, but I agree with
>>>>you in principle. If the image is out-of-focus, it is
>>>>out-of-focus. Period, end of discussion.
>>>
>>>That is not correct. With similar techniques as were used
>>>to restore Hubble Space Station's initial imagery, it is
>>>also possible to restore some of the OOF information.
>>
>> With all due respect to someone I don't know, my first
>> reaction to your contradiction of my simple statement is,
>> well, horseshit.
>
> Bart is correct. The technology to improve focus is called
> image deconvolution, or image restoration and has been a
> topic of research for decades, well before the Hubble
> problem. One of the more successful algorithms is
> Richardson-Lucy Image Restoration. Photoshop's tools, like
> unsharp mask do not actually sharpen, they only change
> accutance. A combination of edge detection and unsharp
> masking, a method developed by Bob Atkinson, can come close
> to equal it however (I'll be adding this to my web page
> sometime; test done by Bill Hilton).

Bart is /theoretically/ correct. Go back and re-read my last
paragraphy. I don't give a tinker's damn about what can be
done in a lab with hours of effort using expensive software.

The issue being discussed was how to save a badly out-of-focus
image and not what the NASA people did or what esoteric
software can do under carefully controlled conditions on
special case problems.

Photography used to be both art and science, but more art from
which the "science" was built around. Today, particularly in
this NG, debates like this almost immediately lose their
relavance to the original innocent request for help into an
elitist theoretical debate. If that's what you want to do,
fine by me.

> See:
> Image Restoration
> Using Adaptive Richardson-Lucy Iteration
> http://www.clarkvision.com/imagedeta...e-restoration1
>
> Image Restoration Using the Damped Richardson-Lucy Method
> http://www.stsci.edu/stsci/meetings/...dings/whiter_d
> amped.dir/whiter_damped.html
>
> There is no RL pluggin for Photoshop that I am aware of.
> Bart and I both use an image processing program called
> ImagesPlus which has this and other image restoration
> algorithms.
>
> Basically, the concept is this: consider a blurred image,
> e.g. due to focus, or motion. Adjacent pixels have image
> information of other pixels. Using a model of the blur,
> the software estimates the contribution of the blur to each
> pixel and moves that signal back to adjacent pixels. It is
> an iterative process and takes a lot of computation, but
> can work very well. But there is no free lunch. The
> process increases noise, and can cause ringing artifacts if
> done to extreme, and/or if the blur model doesn't match the
> image.
>
> In my own experiments, it seems there is roughly an even
> trade in noise versus resolution. My digital workflow now
> includes RL restoration on any image I intend to print
> large. I routinely double the pixel count in each
> dimension and produce very sharp large prints (e.g. 16x24
> inches) from 8-megapixel images.
>
Do you do this for real or for an intellectual exercise?
Again, you admit to what I already said - fixing an OOF image
is a compromise between noise and sharpness, and almost always
results in greatly increased artifacts. Actually, the
artifacts and noise were already there, they were simply
exacerrbated by the attempted sharpness/detail illusory
process.

--
ATM, aka Jerry

"Whether You Think You CAN Or CAN'T, You're Right." – Henry
Ford

Today Dave Martindale commented courteously on the subject at
hand

>>Today Bart van der Wolf commented courteously on the
>>subject
>>> That is not correct. With similar techniques as were used
>>> to restore Hubble Space Station's initial imagery, it is
>>> also possible to restore some of the OOF information.
>
> All Things Mopar <> writes:
>>With all due respect to someone I don't know, my first
>>reaction to your contradiction of my simple statement is,
>>well, horseshit.
>
> Maybe, but you're wrong. Try searching for
> "deconvolution".
>
>>What did it cost to fix Hubble, several /billion/ dollars?
>>And, they did /not/ fix the out-of-focus images /after/
>>they'd been downloaded to earth, those are toast (yes, they
>>tried and tried and tried, but never succeeded to any
>>measurable degree).
>
> They did succeed in sharpening many images in the several
> years before the Hubble optics were fixed. The Hubble was
> unable to shoot images of dim stars, one thing it was
> originally designed for, because the light was spread over
> so many pixels. But bright objects could be imaged and the
> defocus largely corrected, because they knew the
> telescope's blur function so well.

Again, I say "horseshit". If they succeeded so well, why did
NASA spend /years/ and countless /millions/ of dollars and
mount a very specialized shuttle mission to correct the
Hubble's underlying problem if they could simply correct it's
blurry images using a technique as simple as deconvolution?

You theoretical elitists have to understand an indisputable
fact: pixels cannot be created where there are none. When
looking closely at the pixel map of a sharp, detailed, in-
focus image vs. a blurry, obviously out-of-focus image, the
pixels defining the fine detail are simply not there and no
amount of hand waving and black magic can bring them back. All
that can be done is various mathematically valid ways to trick
the human eye into /thinking/ the image is sharper.

> And yes it cost a lot to fix Hubble, but that was a
> hardware fix. One camera module was replaced by a very
> complicated mirror assembly that corrected the error in the
> primary mirror. That doesn't mean the software sharpening
> of the blurred images cost this much.
>
>>NASA engineers and astronomers colaborated to add software
>>to Hubble and some electronics (as I recall, but I'm no
>>Hubble expert) to compensate - not correct - the incorrect
>>mirror curvature grind to allow /new/ images to be
>>in-focus, and then through some pretty sophisticated -
>>read: extremely expensive mathematical techniques.
>
> You're just making this up, aren't you?
>
> No, the correction device was hardware only. Once
> installed, the telescope delivered sharp images to the
> various sensors without any postprocessing needed. As I
> understand it, the added optics were arranged to image the
> primary mirror onto a special correction mirror. Because of
> this, all the light that reflected from a particular point
> on the primary mirror was re-gathered at a single point on
> the correction mirror. That allowed the error in the
> primary mirror to be corrected by a complementary
> correction in the new mirror. All the light from a single
> star was focused on a tiny spot a few pixels in size, as
> designed.
>
>>And, the space inside Hubble the
>>astronaut had to work in was so tight and the chance to
>>irreparably damage the telescope was so high that nobody
>>really knew in advance if Hubble could or could not be
>>fixed.
>
> Really, what's this got to do with software sharpening?
>
>>> The result will not be perfect, because the truely lost
>>> information will generate artifacts during the
>>> restoration process. It is also important to have a good
>>> model for the de-focus. Some so-called "blind"
>>> deconvolution algorithms estimate the blur function,
>>> other methods require prior input of the model to be
>>> used.
>
>>What kind of techno babble is this?
>
> It makes perfect sense if you'd bother to do any kind of
> search for deconvolution methods.

I know what deconvolution is but my statement stands: how can
a person blowing their vacation pictures predict just how it
was done so they can give the correct model to the software?
Further, how much does it cost, how is it used, how long per
image does it take to optimize results, and what nasty side-
effects are created?

>>I'm
>>supposed to know in advance exactly how I blew it, so the
>>software knows how to begin? If I'm misunderstanding you, I
>>apologize but I fail to see how someone can predict the
>>way(s) to blow a focus.
>
> It could be as simple as having a bright point source in
> the image (the reflection of the sun off a small metal
> ball, for example) which allows estimating exactly what the
> blur function looks like. If you know what the blur did to
> the image, you can invert it (with the caveats that Bart
> mentioned).

I seriously doubt that a few spots of glare across an
otherwise bland expanse of pixels typical of a blurry image
are enough for even sophisticated software to latch onto and
figure out how to reconstruct hundreds of thousands - or
millions - of missing pixels.

> As you point out, all this may not be worthwhile in most
> cases in general photography. But you as much as called
> Bart a liar, while at the same time mangling much of what
> you said about Hubble.
>
I never used the word "liar" but if you want to interpret my
comments that way, that's your privelige. I also said I'm no
Hubble expert, so I'd hardly call what I said "mangling" it.
And, your 1st statement, where you agree with my thesis, says
it all: these techniques are /not/ for the general case, so
have no general use whatsoever to real-world photographers
trying to save their European vacation pictures or pictures of
their daughters wedding.

You elitists need to get out of the basements where the
computer wonks live and come up and smell the fresh air where
normal people live and discuss /in practical terms/ how to use
these esoteric techniques in a reasonable inexensive, time-
efficient, least side-effect way on /real/ photo image
situations across a large spectrum of subjects and lighting
conditions. Until/unless you can do that, when I smell
unadulterated crap, I'll yell "horseshit".

Bye!

--
ATM, aka Jerry

"Whether You Think You CAN Or CAN'T, You're Right." – Henry
Ford

Today ASAAR commented courteously on the subject at hand

> On Tue, 21 Mar 2006 20:59:54 -0600, All Things Mopar wrote:
>
> Today All Things Mopar once again showed his inability to
> comment courteously on the subject at hand:
>
>>> That is not correct. With similar techniques as were used
>>> to restore Hubble Space Station's initial imagery, it is
>>> also possible to restore some of the OOF information.
>>
>> With all due respect to someone I don't know, my first
>> reaction to your contradiction of my simple statement is,
>> well, horseshit.
>
> With all due respect, you've once again shown yourself to
> be a
> horse's ass. There are a number of people here that know
> far more than you about many areas of photography, and Bart
> is one of them. But you wouldn't know that, as you seem to
> value typing much more highly than reading and
> comprehension.
>

Well, you've got this right, I /am/ a horse's ass. But, at
least, I provide /practical/ information, whereas you simply
exist to lurk and lay in wait to pounce on someone you don't
like. What exactly have you contributed to this thread except
to insult me personally?

I'm aware that people here know more about photography than
me; that isn't hard to do. But, if they cannot transmit their
knowledge to people asking for help, their knowledge is
useless. As for you, I doubt you know anything at all except
that you're skilled at putting people down.

As to typing, reading, and comprehension, yes, I practice
Mavis Beacon so I can type drivel at high rates of speed. And,
what would you know about my reading and comprehension
abilities? Why don't /you/ enlighten me and the others? Oh,
you can't do that, as you're nothing but a bullshit artist
with delusions of grandeur.

>> Before you or someone else decides to take me on about
>> this, please keep this in mind: I am a pragmatist, not a
>> theorotician. I am also not an elitest. I deal in reality.
>
> Your reality is often the purest fantasy.

That is your interpretation. It is wrong, but you can live in
your own version of reality or fantasy if that is your
choosing. People who actually have their feet planted in the
real world would be able to quantiatitively pick apart your
logic - if there were any. You never actually provide
information, as I said above, you just spew insults using
conclusions based on incorrect and unfounded assumptions.
>
>> Two sayings come to mind here - "the proof of the pudding
>> is in the eating" and "I don't know anything about art,
>> but I know what I like.
>
> You've evidently developed a taste for lip-smackin BS.

No, but I like smacking trolls like you. Here's another saying
for you to chew on when you go back under the bridge with the
other trolls to sleep during the day - "I don't feed trolls, I
feed /ON? trolls".
>
>> Do whatever floats your boat and I'll do the same. And,
>> you have a good evening, hear?! <grin>
>
> And I'll leave you to smile as your boat floats down that
> well
> known river immortalized by the Fugs. <grin and bear it>
>
Smile or frown, don't make never mind to me. If you want to
debate this or any other subject you and I have tangled on at
some factual, intelligent level, I'd be happy to debate it
with you. However, I have yet to see you say anything factual
or useful on /any/ subject, so I see no purpose to listening
to or talking to you again.

--
ATM, aka Jerry

"Whether You Think You CAN Or CAN'T, You're Right." – Henry
Ford

All Things Mopar wrote:
> Today Roger N. Clark (change username to rnclark) commented
> courteously on the subject at hand
>
>
>>All Things Mopar wrote:
>>>With all due respect to someone I don't know, my first
>>>reaction to your contradiction of my simple statement is,
>>>well, horseshit.
>>
>>Bart is correct. The technology to improve focus is called
>>image deconvolution, or image restoration and has been a
>>topic of research for decades, well before the Hubble
>>problem.

> Bart is /theoretically/ correct. Go back and re-read my last
> paragraphy. I don't give a tinker's damn about what can be
> done in a lab with hours of effort using expensive software.

No, Bart is actually correct. The software is not expensive, in
fact you can get it for free if you want. Commercial software
that does it is more than 3 times less expensive than photoshop.

> The issue being discussed was how to save a badly out-of-focus
> image and not what the NASA people did or what esoteric
> software can do under carefully controlled conditions on
> special case problems.

And I'm not talking about what "NASA people" did. Image restoration
methods are being used by hundreds if not thousands of
terrestrial photographers and astrophotographers every day.

> Photography used to be both art and science, but more art from
> which the "science" was built around. Today, particularly in
> this NG, debates like this almost immediately lose their
> relavance to the original innocent request for help into an
> elitist theoretical debate. If that's what you want to do,
> fine by me.

The only debate here is knowledgeable people have pointed out
that this technology exists and works well, and you seem
to argue that it doesn't exist except in theory in a lab.
Try reading the links. If you want links to many images
by many people who do this routinely, I can supply some.
There really are some awesome results coming out
of the technology.

>>See:
>> Image Restoration
>> Using Adaptive Richardson-Lucy Iteration
>> http://www.clarkvision.com/imagedeta...e-restoration1
>>
>>Image Restoration Using the Damped Richardson-Lucy Method
>>http://www.stsci.edu/stsci/meetings/...dings/whiter_d
>>amped.dir/whiter_damped.html

> Do you do this for real or for an intellectual exercise?
> Again, you admit to what I already said - fixing an OOF image
> is a compromise between noise and sharpness, and almost always
> results in greatly increased artifacts. Actually, the
> artifacts and noise were already there, they were simply
> exacerrbated by the attempted sharpness/detail illusory
> process.

Again, the technology is real and being used by
many photographers. Yes, I do this for real.
I have sold many photographs that have been processed
with RL restoration. The same argument of greatly increased
artifacts can be said of almost any tool, whether simple curves
or levels, unsharp mask, of photoshop's newer
smart sharpen (which again does not sharpen, only increases
accutance). So like any tool, there is skill in using
it to produce good results.

Roger

All Things Mopar wrote:

> Today Dave Martindale commented courteously on the subject at
> hand
>
>

> Again, I say "horseshit". If they succeeded so well, why did
> NASA spend /years/ and countless /millions/ of dollars and
> mount a very specialized shuttle mission to correct the
> Hubble's underlying problem if they could simply correct it's
> blurry images using a technique as simple as deconvolution?

It is very simple. Scientific investigations usually require high
signal-to-noise ratios to measure specific properties.
While the deconvolution of Hubble images improved spatial
image quality, it did so at the expense of increased noise.
Then a faint source was spread over many pixels and could be
lost in the noise of the sensor. That noise also
made some detections of faint signals next to bright
objects impossible. The optical fix solved the problem well.

> You theoretical elitists have to understand an indisputable
> fact: pixels cannot be created where there are none. When
> looking closely at the pixel map of a sharp, detailed, in-
> focus image vs. a blurry, obviously out-of-focus image, the
> pixels defining the fine detail are simply not there and no
> amount of hand waving and black magic can bring them back. All
> that can be done is various mathematically valid ways to trick
> the human eye into /thinking/ the image is sharper.

No. There is information there. The information is just
spread over a number of pixels. It is not theoretical,
and no hand waving and black magic is involved, unless
you consider mathematics black magic; I don't.


> I know what deconvolution is but my statement stands: how can
> a person blowing their vacation pictures predict just how it
> was done so they can give the correct model to the software?
> Further, how much does it cost, how is it used, how long per
> image does it take to optimize results, and what nasty side-
> effects are created?

If you read papers and web sites doing some simple google
research, you can find the answers. I gave you two
links, but you apparently didn't read them.


> I seriously doubt that a few spots of glare across an
> otherwise bland expanse of pixels typical of a blurry image
> are enough for even sophisticated software to latch onto and
> figure out how to reconstruct hundreds of thousands - or
> millions - of missing pixels.

The pixels are not missing; they are still the same ones
in the original image, just the information is spread out
over many pixels. You can examine fine detail, estimate the
width of an edge and make a close approximation of the
blur.


> my thesis, says
> it all: these techniques are /not/ for the general case, so
> have no general use whatsoever to real-world photographers
> trying to save their European vacation pictures or pictures of
> their daughters wedding.

This is like saying unsharp mask is not for real-world
photographers. Simply because you don't know the technology
doesn't mean it doesn't exist and doesn't work.
>
> You elitists need to get out of the basements where the
> computer wonks live and come up and smell the fresh air where
> normal people live and discuss /in practical terms/ how to use
> these esoteric techniques in a reasonable inexensive, time-
> efficient, least side-effect way on /real/ photo image
> situations across a large spectrum of subjects and lighting
> conditions. Until/unless you can do that, when I smell
> unadulterated crap, I'll yell "horseshit".

I think you are the one being the elitist.
Image restoration is being done, it can be done for free (software)
or modest cost (<$200). Fast computers these days mean pretty
fast results, but longer than unsharp mask (I don't have
a very fast computer and still use it). I and others
use the RL algorithms routinely and produce images
that are sharper, and i make larger prints that sell.

Roger

On Wed, 22 Mar 2006 06:13:51 -0600, All Things Mopar
<> wrote:

>Again, I say "horseshit". If they succeeded so well, why did
>NASA spend /years/ and countless /millions/ of dollars and
>mount a very specialized shuttle mission to correct the
>Hubble's underlying problem if they could simply correct it's
>blurry images using a technique as simple as deconvolution?

Hubble's underlying problem is with its main mirror, and it wasn't
fixed. Instead, NASA installed corrective mirrors.
Until those mirrors could be designed, made and installed,
deconvolutiuon techniques were used (and very well, BTW) to get good
images.
Why use the mirrors? Because the software techniques weren't as good
as the correctibe mirrors.
This doesn't mean the software solution isn't good, but that it isn't
as good as the hardware solution.
But the deconvolution technique did work, and it continues to do so
today.

--
Bill Funk
replace "g" with "a"

One can "sharpen" a picture, but that is not the same as fixing a focus
problem. A focus problem can be that only some objects in the scene
are blurry, not the whole picture. This kind of problem is very hard
to fix.

Most routines that perport to fix a focus blur apply a sharpening
filter to the whole image. This means that if there were parts of the
image that WERE in focus already, they will be oversharpened.

These filters work okay if the WHOLE image is blurry, but not if only
part is out of focus.

Oversharpening leads to noise and artifacts similar to JPEG artifacts.