Reading mode: Light Dark

Foveon/Bayer comparison

Go to last post
sg10 said:
Dominic Groß said:
First of all the most important point remains, the moire is a
luminace artifact.
Click to expand...
We'll get to that if we can ever get past the first point, which is
that JPEG'ing, without a doubt, destroys the camera's colorspace...
Click to expand...
JPEG maintains any color space you care to tag the image with. You can JPEG encode an image that's in an sRGB color space, and it will remain in SRGB, with a delta E due to the JPEG encoding of less than 0.04, quite impossible to see. You can just as easily JPEG encode an image in the Adobe RGB color space, or CIELAB, for that matter, and it will decode into the proper color space. With a little work, you can even JPEG encode in the camera's raw sensor color space.

Color space issues aside, the higher quality JPEG modes degrade images less through coding losses than the initial capture noise of a digital camera does.

I realize you do not have my background, you pbviously have no formal psychophysics, visual science, or math education, but these are not hard terms to look up. For the sake of the group which you are damaging by your behavior, please do a little research before posting further.

--
Ciao!

Joe

http://www.swissarmyfork.com
 
sg10 said:
The fact is, JPEG reduces the colorspace so dramatically and adds
so many aritifacts that there is no possible way to "analyze"
subtleties in any image, let alone supposed moire in an SD-9 image.
Most savvy discussions about the effect some say they see under
certain rare conditions is nothing more than a manifestation of the
display media's raster pattern. In fact, if you look at a
"perfect" single-pixel B&W checkboard pattern on your monitor right
now, you'll see clear rainbowing thoughout. That's not SD-9 moire
either.
Click to expand...
The fact is that you have no idea what you're talking about. You don't understand colorspaces or jpeg compression. You're just making stuff up.
sg10 said:
  • 36-bit data is was already wiped out, only 1/4096th of the
colorspace is there to "analyze."
Click to expand...
You haven't shown that the additional information is visible to humans on any known display device.
sg10 said:
  • Of the remaining 400K colors (avg-ish), from the orignal 3M
(avg-ish) 36-bit image, about half of those are eliminated and the
result is dithered to simulate the residue from of the original
image after conversion to down to 16.8M color choices from the
original 68 Billion.
Click to expand...
Repeating things known to be false is lying. You've been told that there is nothing dithering but you keep repeating it. Stop lying. Learn about color and then try to say something intelligent about what jpeg is doing.
sg10 said:
  • When JPEG'ing a JPEG (even once), the unique color count usually
goes up, not down. A lot. This isn't new optical data, it is
statistical noise introduced largely in algorithmic/harmonic color
patterns.
Click to expand...
You've been told already that jpeg is deterministic. There is no "statistical noise."

There is no difference between adding undetectable color changes and taking them away. You seem to have no clue about compression levels and their relation to human vision.
sg10 said:
  • My B&W image showed this vividly using only 2 colors, a JPEG of
an SD-9 image is doing its number on around 200K unique colors of
the 400K that remained from the 3M originals. So, about 100,000x
the color variation. A JPEG of a JPEG is taking those and adding
about double the number of unique colors back again (this is noise,
not data), and doing the same with what data remains, plus acting
on noise.
Click to expand...
And this and your other nuggets demonstrate absolutely nothing.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/
 
sg10 said:
Are you seriously claiming there is no difference between 36-bit
and 24-bit colorspace? Again?

I thought we ironed this out a month ago? You don't remember this...?
Click to expand...
We did iron it out, and we do remember it. It's just that you are the only person who remembers anyone agreeing with you.

The JPEG file original.jpg is in a full 24 bit representation (not a color space). It is tagged with the Adobe RGB (1998) color space. It has 8 bits each for red, green, and blue. On any 24 bit video system, using software that properly decodes JPEG into a 24 bit representation it displays smoothly, without visible banding. This includes everything from well refined commercial programs like Adobe PhotoShop to the freeware IrfanView or GIMP.

Most browsers, including Internet Explorer, Netscape, or Mozilla, display JPEG files using a fast 16 bit JPEG decompression, into a 5:6:5 representation. They only allocate 5 bits (32 distinct shades) to blue. So displaying your picture in a browser results in noticible banding.

This banding is not moire. The moire in the earlier examples is clearly visible no matter what software the image is viewed in.



--
Ciao!

Joe

http://www.swissarmyfork.com
 
Joseph S. Wisniewski said:
Joseph S Wisniewski said:
Are you seriously claiming there is no difference between 36-bit
and 24-bit colorspace? Again?

I thought we ironed this out a month ago? You don't remember this...?
Click to expand...
We did iron it out, and we do remember it. It's just that you are
the only person who remembers anyone agreeing with you.
Click to expand...
We should collect the links to all this somewhere to have it at hand in situations like this one...

--
Regards from Old Europe,

Dominic

http://www.pbase.com/sigmasd9/dominic_gross
 
Ron Parr said:
sg10 said:
The fact is, JPEG reduces the colorspace so dramatically and adds
so many aritifacts that there is no possible way to "analyze"
subtleties in any image, let alone supposed moire in an SD-9 image.
Most savvy discussions about the effect some say they see under
certain rare conditions is nothing more than a manifestation of the
display media's raster pattern. In fact, if you look at a
"perfect" single-pixel B&W checkboard pattern on your monitor right
now, you'll see clear rainbowing thoughout. That's not SD-9 moire
either.
Click to expand...
The fact is that you have no idea what you're talking about. You
don't understand colorspaces or jpeg compression. You're just
making stuff up.
Click to expand...
For your education...

http://www.google.co.kr/search?hl=ko&ie=UTF-8&oe=UTF-8&newwindow=1&q=moire+shadow-mask+dot-pitch+mismatched+pixels&lr=

You've added nothing new below. You are simply ignoring irrefutable proof as you are clearly out of ideas and don't know how to apologize.
Member said:
Member said:
  • 36-bit data is was already wiped out, only 1/4096th of the
colorspace is there to "analyze."
Click to expand...
You haven't shown that the additional information is visible to
humans on any known display device.
Member said:
  • Of the remaining 400K colors (avg-ish), from the orignal 3M
(avg-ish) 36-bit image, about half of those are eliminated and the
result is dithered to simulate the residue from of the original
image after conversion to down to 16.8M color choices from the
original 68 Billion.
Click to expand...
Repeating things known to be false is lying. You've been told that
there is nothing dithering but you keep repeating it. Stop lying.
Learn about color and then try to say something intelligent about
what jpeg is doing.
Member said:
  • When JPEG'ing a JPEG (even once), the unique color count usually
goes up, not down. A lot. This isn't new optical data, it is
statistical noise introduced largely in algorithmic/harmonic color
patterns.
Click to expand...
You've been told already that jpeg is deterministic. There is no
"statistical noise."

There is no difference between adding undetectable color changes
and taking them away. You seem to have no clue about compression
levels and their relation to human vision.
Member said:
  • My B&W image showed this vividly using only 2 colors, a JPEG of
an SD-9 image is doing its number on around 200K unique colors of
the 400K that remained from the 3M originals. So, about 100,000x
the color variation. A JPEG of a JPEG is taking those and adding
about double the number of unique colors back again (this is noise,
not data), and doing the same with what data remains, plus acting
on noise.
Click to expand...
And this and your other nuggets demonstrate absolutely nothing.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/
Click to expand...
--
http://www.pbase.com/imageprocessing/sd9
 
Joseph S. Wisniewski said:
Joseph S Wisniewski said:
Are you seriously claiming there is no difference between 36-bit
and 24-bit colorspace? Again?

I thought we ironed this out a month ago? You don't remember this...?
Click to expand...
We did iron it out, and we do remember it. It's just that you are
the only person who remembers anyone agreeing with you.

The JPEG file original.jpg is in a full 24 bit representation
Click to expand...
...of a 36-bit image.
Joseph S Wisniewski said:
(not
a color space). It is tagged with the Adobe RGB (1998) color space.
It has 8 bits each for red, green, and blue. On any 24 bit video
system, using software that properly decodes JPEG into a 24 bit
representation it displays smoothly, without visible banding. This
includes everything from well refined commercial programs like
Adobe PhotoShop to the freeware IrfanView or GIMP.

Most browsers, including Internet Explorer, Netscape, or Mozilla,
display JPEG files using a fast 16 bit JPEG decompression, into a
5:6:5 representation. They only allocate 5 bits (32 distinct
shades) to blue. So displaying your picture in a browser results in
noticible banding.
Click to expand...
A smooth gradient using 24-bit color will band using any program. There simply are not enough color choices to avoid it. That's why the standard was replaced.
Joseph S Wisniewski said:
This banding is not moire.
Click to expand...
Color banding is color banding. What does moire have to do with having only 256 shades per channel? You are confusing several discrete issues.
Joseph S Wisniewski said:
The moire in the earlier examples is
clearly visible no matter what software the image is viewed in.



--
Ciao!

Joe

http://www.swissarmyfork.com
Click to expand...
--
http://www.pbase.com/imageprocessing/sd9
 
ding1 said:
sg10 said:
Yi Ding said:
At what JPEG quality did you save the images? I don't have PS on
the machine I'm using now, so I saved the PNG file (resized to 25%
so as to match the size of the original file you worked on) three
times in MS PhotoEditor at 100% quality,
Click to expand...
You are confusing a compression parameter with a quality %. 100 is
a parameter, it is not a quality percentage nor was it ever meant
to be associated, or proportional in any way to one.
Click to expand...
Well, it was my mistake that I added a "%" mark to the number,
which is called "JPEG quality factor" in MS PhotoEditor that ranges
from 1 to 100. However, the fact that I have an error doesn't mean
you are correct in your claims. This is again a basic logic.
Click to expand...
So you plead, "don't leave it on a slam dunk without repsonding to my questions," then you respond to nothing I said?
 
sg10 said:
No, everyone knows the effect of pouring 8000 gallons into a 1
gallon bucket causes unwanted artifacts.
Click to expand...
You can call quantization error an "artifact" if you like, I won't quibble over a non-mathematical term, but it is not of the kind of error which causes Moire.

Please stop changing the subject!

The subject at hand is your claim that JPEG compression causes Moire. You haven't proven it yet. Since your claim is close to being libelous of the JPEG work, this matter is of some importance to many.

Please address your Moire claim before you move on to your next claim. I have given you a URL to get free, real live code that you can delve into to substantiate your claim. If you'd lost the link, I can repeat it.
  • kc
 
sg10 said:
A smooth gradient using 24-bit color will band using any program.
There simply are not enough color choices to avoid it. That's why
the standard was replaced.
Click to expand...
Sounds like you need to buy a better monitor or graphics card if you're having these problems.
sg10 said:
sg10 said:
This banding is not moire.
Click to expand...
Color banding is color banding. What does moire have to do with
having only 256 shades per channel? You are confusing several
discrete issues.
Click to expand...
Hmmm... Could it be that your your first factually correct statement in days is buried in the above?

Color banding, real or imagined, indeed has nothing to do with moire. The sad thing is, you're the one who brought it up in your effort to explain away basic sampling theory.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/
 
Ron, I don't understand why you keep playing straight man for this buffoon.

He is clearly not interested in learning anything, or ever conceding a point, or even holding a sensible dialogue. He will continue to get the last word so it is better to let to just walk away and let him have the last word once instead of ten times.

The last time a thread got this big I think he got up at 4 AM, to respond 6 times and close it out so he could be sure to get the last word in. He is clearly obsessed.

Peter
Ron Parr said:
sg10 said:
You've added nothing new below. You are simply ignoring
irrefutable proof as you are clearly out of ideas and don't know
how to apologize.
Click to expand...
Irrefutable proof of what? You don't even understand the words
you're using well enough to explain your thesis.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/
Click to expand...
--
http://www.trytel.com/~pguidry/vacation.html
 
Peter G said:
Ron, I don't understand why you keep playing straight man for this
buffoon.
Click to expand...
Perhaps it's unrealistic, but I'd like to think that the concerted efforts of several informed people can subvert the efforts of one man to spread misinformation and lies.
Peter G said:
The last time a thread got this big I think he got up at 4 AM, to
respond 6 times and close it out so he could be sure to get the
last word in. He is clearly obsessed.
Click to expand...
I noticed that too...

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/
 
Somebody has to do that job, can you imagine what happens if people believe this stuff?

And an even bigger problem is what Joe pointed out below about the quality of this forum...

http://forums.dpreview.com/forums/read.asp?forum=1027&message=6166139
Peter G said:
He is clearly not interested in learning anything, or ever
conceding a point, or even holding a sensible dialogue. He will
continue to get the last word so it is better to let to just walk
away and let him have the last word once instead of ten times.

The last time a thread got this big I think he got up at 4 AM, to
respond 6 times and close it out so he could be sure to get the
last word in. He is clearly obsessed.

Peter
Ron Parr said:
sg10 said:
You've added nothing new below. You are simply ignoring
irrefutable proof as you are clearly out of ideas and don't know
how to apologize.
Click to expand...
Irrefutable proof of what? You don't even understand the words
you're using well enough to explain your thesis.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/
Click to expand...
--
http://www.trytel.com/~pguidry/vacation.html
Click to expand...
--
Regards from Old Europe,

Dominic

http://www.pbase.com/sigmasd9/dominic_gross
 
Peter G said:
It depends highly on the strength of your AA filter. You have a D60
I believe. So please show me your worse example in all of your
shots thus far. Even if you do find it the situation is further
improved with modern cameras.

I have an S400 right now and have yet to see ANY in well over a
1000 images.

Canons latest image processing has pretty much eliminated it in the
10D:
Click to expand...
Of course it's hard to find examples when most 10D pictures have been shrunk by a factor of four or more before posting. But here are some that you can detect some moire through the blur:

http://www.pbase.com/image/21232635
A bit around the tip of the eyelashes, and I think the top band of hair.

http://www.pbase.com/image/21232638
Here more where hair crosses over hair, especially at the top of the head.

It's subtle color moire, to be sure - but there is some to be had. Mostly around hair, or at least that's where I notice it most. I don't deny that the SD9 can exhibit luminance moire but such an effect is less distracting that seeing colors pop up where they should not.

--
---> Kendall
http://www.pbase.com/kgelner
http://www.pbase.com/sigmasd9/user_home
 
Ron Parr said:
Sounds like you need to buy a better monitor or graphics card if
you're having these problems.
Click to expand...
Better display fidelity makes all of the below worse, not better.
Member said:
Color banding, real or imagined, indeed has nothing to do with
moire. The sad thing is, you're the one who brought it up in your
effort to explain away basic sampling theory.
Click to expand...
Right, I'm the one who said non-interpolated images sample greens at too low a green frequency to guess their colors with haigh enough accuracy to avoid moire on a 24-bit JPEG with 1/4096th the original 36-bit colorspace then reduced by half the unique colors and dithered to compensate before it was JPEG'd again to piggyback on the original artifacts and double the unique color count back up to near the original while adding zero optical information and while most pixels are slightly misaligned to the shadow-mask of the monitor which will moire even a perfectly lossless image which displays single pixel resolution.
 
sg10 said:
Ron Parr said:
Sounds like you need to buy a better monitor or graphics card if
you're having these problems.
Click to expand...
Better display fidelity makes all of the below worse, not better.
Click to expand...
Erm... The appearance of banding in 24 bit images can only be explained by faulty hardware or software.
sg10 said:
sg10 said:
Color banding, real or imagined, indeed has nothing to do with
moire. The sad thing is, you're the one who brought it up in your
effort to explain away basic sampling theory.
Click to expand...
Right, I'm the one who said non-interpolated images sample greens
at too low a green frequency to guess their colors with haigh
enough accuracy to avoid moire on a 24-bit JPEG with 1/4096th the
original 36-bit colorspace then reduced by half the unique colors
and dithered to compensate before it was JPEG'd again to piggyback
on the original artifacts and double the unique color count back up
to near the original while adding zero optical information and
while most pixels are slightly misaligned to the shadow-mask of the
monitor which will moire even a perfectly lossless image which
displays single pixel resolution.
Click to expand...
Do you expect anybody to understand the above?

http://www.bartleby.com/141/

Study. Try again.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/
 
Ron Parr said:
sg10 said:
Ron Parr said:
Sounds like you need to buy a better monitor or graphics card if
you're having these problems.
Click to expand...
Better display fidelity makes all of the below worse, not better.
Click to expand...
Erm... The appearance of banding in 24 bit images can only be
explained by faulty hardware or software.
Click to expand...
Absurd, why do you suppose 24-bit capture devices (cameras/scanners) barely qualify as prosumer, entry level quality. Some sites like this one claim that even 48-bit color is sometimes not enough, I'd say that is an overstatement.

http://www.panoscan.com/HDR.html

Conversely, I'd say that their statement that 24-bit can be acceptable under some conditions is equally wrong. 24-bit is never sufficient for professional work, unless 1990's era bandwidth requirements are the limiting factor, and sadly this is often the case.
Member said:
Member said:
Member said:
Color banding, real or imagined, indeed has nothing to do with
moire. The sad thing is, you're the one who brought it up in your
effort to explain away basic sampling theory.
Click to expand...
Right, I'm the one who said non-interpolated images sample greens
at too low a green frequency to guess their colors with haigh
enough accuracy to avoid moire on a 24-bit JPEG with 1/4096th the
original 36-bit colorspace then reduced by half the unique colors
and dithered to compensate before it was JPEG'd again to piggyback
on the original artifacts and double the unique color count back up
to near the original while adding zero optical information and
while most pixels are slightly misaligned to the shadow-mask of the
monitor which will moire even a perfectly lossless image which
displays single pixel resolution.
Click to expand...
Do you expect anybody to understand the above?
Click to expand...
Why you think such a workflow is worth spending thousands on a DSLR and even more on lenses? No. I don't.
 
sg10 said:
Conversely, I'd say that their statement that 24-bit can be
acceptable under some conditions is equally wrong. 24-bit is never
sufficient for professional work, unless 1990's era bandwidth
requirements are the limiting factor, and sadly this is often the
case.
Click to expand...
I was actually thinking JPEG there, when I said "24-bit." 24-bit .png/.tif/lossless is on the margin.
 
Kok Chen said:
sg10 said:
No, everyone knows the effect of pouring 8000 gallons into a 1
gallon bucket causes unwanted artifacts.
Click to expand...
You can call quantization error an "artifact" if you like, I won't
quibble over a non-mathematical term, but it is not of the kind of
error which causes Moire.

Please stop changing the subject!
Click to expand...
You're the ones saying the SD-9 pic is displaying moired, I'm saying its not. Now it seems you are agreeing with me, wonderful.
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top