Fujifilm claims resolutions of almost double their ccd pixel count
Uncle Frank
theme; they claimed an effective resolution of about twice the pixel
count of their ccd. As an example, from their S602Z announcement,
"New 3rd Generation Super CCD sensor and advanced LSI algorithm (3.1
million effective pixels) producing class-leading 6.03 million
recorded pixels (2,832 x 2,128)".
Is "recorded pixels" a meaningful measure, or slight of hand. I get
the impression they may be doing something like onboard cubic
interpolation to increase pixel count, but you could do the same
thing with any picture file with Photoshop.
uf
Graham Sumner
of the reasons why I went for an E10 rather than
Fuji's SLR, despite really liking their cameras. By
their measure, using interpolation the E10 is an
8mp camera.
Graham
"Uncle Frank" <uncle...@hotmail.com> wrote in message
news:df512813.02013...@posting.google.com...
Todd Walker
uncle...@hotmail.com says...
You are exactly right -- this is an interpolated resolution. This is
nothing new for Fuji, they have been playing the interpolation game for
years with their digitals. They claim that it is valid because of the
unique design of the SuperCCD. Bullshit. It is dishonest marketing hype,
plain and simple.
--
*-*-*-*-*-*-*
Todd Walker
http://twalker.d2g.com
Canon Pro90IS:
http://twalker.d2g.com/pro90/index.htm
Pbase galleries:
http://www.pbase.com/twalker294
*-*-*-*-*-*-*
RickDecker
the same shot with the3 and 6 mp resolution and printing a large-size
picture and comparing them. One person actually claimed he got more
detail with the 6mp resolution but I am a little skeptical. I have no
complaints about the camera at this time although I haven't shot it
enough to do a detailed analysis. One thing..the colors are as good or
better than the D30 and it doesn't have the shutter lag of the D30 which
I have shot. It is like shooting Velvia film.
Ian Burley
carried out to produce a standard columnar array - even without changing the
pixel count.
Fujifilm insists their doubled resolution output is not interpolation in the
strictest sense, but in my book they are inventing pixels, so it must be
interpolation.
My experience with a FinePix 4800 Zoom is that there is no great benefit in
the double resolution mode, except perhaps if you don't want the extra stage
of interpolating up for bigger than average prints using something like
Photoshop.
Personally, I doubt I'd ever use the interpolated mode.
Having said that, someone has posted a link to a 6MP (double resolution) S1
Pro image on my site's discussion forum and I have to say that despite
considerable compression (the file is only a mite over 1MB), the
photographic quality of the result is very impressive - maybe not in
sharpness, but tone; when printed it really does look like a film shot.
The message in question is at:
http://www.dp-now.com/cgi-bin/forum/forum.pl?read=113
Ian
--
Digital Photography Now
UK-based digital photo Web magazine
www.dp-now.com
"Uncle Frank" <uncle...@hotmail.com> wrote in message
news:df512813.02013...@posting.google.com...
Michael Geary
All consumer digital cameras do this. Fuji just does it more than the
others.
How do you think a 4 megapixel camera gets four million full-color pixels
from a sensor that has one million red subpixels, one million blue
subpixels, and two million green subpixels? It interpolates.
-Mike
Barry Pearson
news:u5i48c...@corp.supernews.com...
>Uncle Frank wrote ...
> > Fujifilm announced 3 new cameras today, and there was a consistent
> > theme; they claimed an effective resolution of about twice the pixel
> > count of their ccd. As an example, from their S602Z announcement,
Be careful about your (Uncle Frank's) use of the word "effective". I suspect
they didn't actually do what you say - "effective" has a specific meaning, and
from what you quote below, they used the word properly, whereas you used it
improperly above.
> > "New 3rd Generation Super CCD sensor and advanced LSI algorithm
> > (3.1 million effective pixels) producing class-leading 6.03 million
> > recorded pixels (2,832 x 2,128)".
> >
> > Is "recorded pixels" a meaningful measure, or slight of hand. I get
> > the impression they may be doing something like onboard cubic
> > interpolation to increase pixel count, but you could do the same
> > thing with any picture file with Photoshop.
>
> All consumer digital cameras do this. Fuji just does it more than the
> others.
>
> How do you think a 4 megapixel camera gets four million full-color pixels
> from a sensor that has one million red subpixels, one million blue
> subpixels, and two million green subpixels? It interpolates.
There are 2 different types of interpolation being discussed. You have just
introduced "colour interpolation", whereas the previous discussion was about
resizing interpolation. They have different characteristics. (Colour
interpolation appears to be very good until the interpolation gets confused when
strong differences in light value occur near the pixel dimensions, then it can
introduce false colours). See:
http://www.extremetech.com/print_article/0,3428,a%253D2383,00.asp
While each of them merits analysis, it is wise to keep them separate. Apart from
anything else, while you can resize in Photoshop, possibly to better effect, as
far as a I know you can't do colour interpolation except in the camera. And it
is desirable to be able to compare 2 cameras - and they "all" (I think) do
colour interpolation.
There is a standard (or proposed standard) for identifying the number of pixels
in a camera, and I think the standard name is "effective number of pixels". This
permits colour interpolation, but excludes what Fuji are doing above. Fuji are
being naughty, or at least pushing the boundaries.
http://www.pima.net/standards/it10/PIMA7000/PIMA700_WD1.0_v3.PDF
--
Barry Pearson
http://www.barrypearson.co.uk/photography/
David Oddie
<ne...@childsupportanalysis.co.uk> wrote:
>There is a standard (or proposed standard) for identifying the number of pixels
>in a camera, and I think the standard name is "effective number of pixels".
This refers to what part of the CCD has the image on it and is largely
down to the way the lens "deposits" light on the CCD.
The extremities of the CCD are just not used because in any camera the
lens is not providing a large enough image to hit the entire sensor.
I am not explaining this very well but I hope you get what I mean.
Essentially the effective number of pixels is an optical
characteristic of the camera and nothing to do with colour or size
interpolation - which as you correctly point out are very different
things.
To justify what Fuji do by saying all cameras do the same (to a lesser
degree) is not correct.
Dave
--
It may be that your sole purpose in life is simply to serve as a
warning to others.
Remove the uppercase N O S P A M to reply via email.
Barry Pearson
news:fi8i5usa0ek6udrlu...@4ax.com...
> On Thu, 31 Jan 2002 10:05:52 -0000, "Barry Pearson"
> <ne...@childsupportanalysis.co.uk> wrote:
>
> >There is a standard (or proposed standard) for identifying the
> >number of pixels in a camera, and I think the standard name
> >is "effective number of pixels".
>
> This refers to what part of the CCD has the image on it and is
> largely down to the way the lens "deposits" light on the CCD.
>
> The extremities of the CCD are just not used because in any camera
> the lens is not providing a large enough image to hit the entire sensor.
>
> I am not explaining this very well but I hope you get what I mean.
Yes, that is what I was saying. (Have you got me confused with someone who
didn't know that?)
> Essentially the effective number of pixels is an optical
> characteristic of the camera and nothing to do with colour or size
> interpolation - which as you correctly point out are very different
> things.
>
> To justify what Fuji do by saying all cameras do the same (to a lesser
> degree) is not correct.
Correct. That is pretty well what I said. I think you are confusing me with
Michael Geary - I was explaining this to him.
Ian Burley
are conventional.
Ian
--
Digital Photography Now
UK-based digital photo Web magazine
www.dp-now.com
"David Oddie" <Dave...@mNaOiSlPaAnMdnews.com> wrote in message
news:fi8i5usa0ek6udrlu...@4ax.com...
DT
anything, is faking pixels by definition. So if you really want any
interpolation at all, I thought it would be better to leave it for you (not
by Fuji) to choose the best software (like PhotoShop, PaintShop or whatever)
to do it.
DT
"Uncle Frank" <uncle...@hotmail.com> wrote in message
news:df512813.02013...@posting.google.com...
David Oddie
<ne...@childsupportanalysis.co.uk> wrote:
>Yes, that is what I was saying. (Have you got me confused with someone who
>didn't know that?)
Pobably - half asleep as usual - but I think I just wanted to be clear
that the effective pixels is an optical thing rather than something to
do with interpolation of any kind or any other factor for that matter.
Barry Pearson
news:kdb68.73056$I8.14...@news4.rdc1.on.home.com...
> Any interpolation, be it from SupperCCD or from your
> Adobe PhotoShop or from anything, is faking pixels by
> definition. So if you really want any interpolation at all,
> I thought it would be better to leave it for you (not
> by Fuji) to choose the best software (like PhotoShop,
> PaintShop or whatever) to do it.
As far as I know, colour interpolation is performed by all (?) digital cameras
(without choice?), but I haven't found the capability in any photo-editing
software. It isn't changing the number of pixels, it is trying to deduce their
RGB values (which can still go wrong, of course).
In other words, be careful with your words "any interpolation" - you may not
really mean it.
DT
get added RGB values or pixels or whatever NOT from the original subject
from which you are taking the photo but from a guess or a calculation. And
this calculation is nothing but interpolation. So my point is nothing but
those added RGB values (as you phrased it) are not from the original
subject.
DT
"Barry Pearson" <ne...@childsupportanalysis.co.uk> wrote in message
news:RVb68.1660$sU.3...@news6-win.server.ntlworld.com...
Don Stauffer
resolution from a digital camera. No one in the consumer digicam
industry that I know uses it. We, the user community, need to push
manufacturers to use this procedure. Mags can help. Whenever they do a
review on a new camera, they should press manufacturers for ISO res
data. They should not publish ANYTHING about 'resolution' unless the
ISO test results are provided. Or, the mags could set up to do the test
themselves.
--
Don Stauffer in Minnesota
stau...@usfamily.net
webpage- http://www.usfamily.net/web/stauffer
Bill Janes
ne...@childsupportanalysis.co.uk says...
>
> As far as I know, colour interpolation is performed by all (?) digital cameras
> (without choice?),
Many digital movie cameras have three CCD's, one for each primary color
and do not perform interpolation. Of course, they mostly have relatively
low resolution.
Bill Janes
Russell Williams
processing to produce a rectangular pixel array since it doesn't start off
that way. Just as the interpolation from a Bayer pattern can produce
much better results than the naive claim of 1/3 the actual number of
pixels on the CCD, the Fuji process should be able to produce
better results than the number of pixels on their CCD. Surely not
2X, but it's not a complete fabrication either. It's also possible that
better algorithms have upped this number in the newest cameras.
If you had the raw sensor data, you could do both the standard Bayer
pattern processing and whatever "interpolation" Fuji is doing in the
computer,
but the data you start with wouldn't look like an image if you put it on
the screen, and the "resize" or "interpolate" commands in image processing
programs would not help, because the processing is really not like
bicubic interpolation.
However, the only way I would use megapixels -- claimed, effective,
or otherwise -- is as a guide for which cameras it would be useful
to compare against. Just look at the resolution results on DPreview
from the DImage 7, Sony F707, and Nikon 5000 -- all of which
reportedly use the same imaging chip. The 707 is as much better
than the D7 as the D7 is better than the best 4MP cameras. A useful
measurement shows how much detail you can actually record through
the lens into the final image.
DPReview reported resolution figures for the 3MP S1: 1300/1200/1000
horizontal / vertical / diagonal lines vs. 900/900/900 for the 3MP Nikon 990
and 1000/1150/1000 for the D30. This is consistent with "not 2X but not
a complete fabrication".
Russell Williams
not speaking for Adobe Systems
Barry Pearson
news:MPG.16c31926...@netnews.attbi.com...
Thanks. I didn't know that. But still cameras?
Ian Burley
these results the way you have will only be meaningful if the same lens was
used, surely?
Ian
--
Digital Photography Now
UK-based digital photo Web magazine
www.dp-now.com
"Russell Williams" <will...@adobe.com> wrote in message
news:Osd68.235$fM1....@newshog.newsread.com...
Tom Holub
Graham Sumner <Graham...@baesystems.com> wrote:
)It's not slight of hand - it's a downright lie and one
)of the reasons why I went for an E10 rather than
)Fuji's SLR, despite really liking their cameras. By
)their measure, using interpolation the E10 is an
)8mp camera.
It is not a lie. The Fuji sensor is hexagonal, and to represent the output
of a hexagonal array on a rectangular output medium (like a computer
screen), you need to use double the pixels. The amount of data is
nearly equivalent to a rectangular sensor of the same size; it is not
a 6-megapixel camera (and Fuji doesn't claim it is).
-Tom
Todd Walker
says...
So you are saying that because of the inferior design of their CCD,
which forces the camera to make up information that is not recorded by
the chip, they should be able to get away with calling their cameras
6mp? I don't think so. It's really pretty simple -- if you capture 3.3mp
worth of information, you have a 3.3mp camera. Rather than playing
semantics and trying to find a loophole that they can exploit, Fuji
needs to be honest with the consumer.
Michael Geary
> > How do you think a 4 megapixel camera gets four million full-color
pixels
> > from a sensor that has one million red subpixels, one million blue
> > subpixels, and two million green subpixels? It interpolates.
Barry Pearson wrote:
> There are 2 different types of interpolation being discussed. You have
just
> introduced "colour interpolation", whereas the previous discussion was
about
> resizing interpolation. They have different characteristics. (Colour
> interpolation appears to be very good until the interpolation gets
confused when
> strong differences in light value occur near the pixel dimensions, then it
can
> introduce false colours). See:
> http://www.extremetech.com/print_article/0,3428,a%253D2383,00.asp
Thanks for clearing that up. That is a very informative article!
-Mike
AM
says...
> What really bugs me is that there is an ISO standard for measuring
> resolution from a digital camera. No one in the consumer digicam
> industry that I know uses it. We, the user community, need to push
> manufacturers to use this procedure. Mags can help. Whenever they do a
> review on a new camera, they should press manufacturers for ISO res
> data. They should not publish ANYTHING about 'resolution' unless the
> ISO test results are provided. Or, the mags could set up to do the test
> themselves.
What is this ISO standard ?
--
Alfred Molon
Remove the obvious to email
------------------------------
Olympus C4040 discussion group at
http://groups.yahoo.com/group/Olympus4040
AM
> It is not a lie. The Fuji sensor is hexagonal,
No it's not. It's a rectangular array rotated by 45 degrees.
Dave Martindale
>There are 2 different types of interpolation being discussed. You have just
>introduced "colour interpolation", whereas the previous discussion was about
>resizing interpolation.
Actually, there are really three different variants of interpolation being
discussed. There is the image processing that generates a 3-colour image
from a single sensor with a mosaic filter on it. There is the processing
that Fuji does to convert their rotated-raster sensor into an image that
can be stored in a conventional image file. And there is simple resizing
interpolation.
Photoshop can do the last of these, so there's little reason to do it
in camera (which is why digital zoom is pretty useless). Fuji's
resampling is related to what Photoshop does, but it's not the same.
And it *has* to be done in camera to avoid losing some information
that the sensor does capture - at least, if the output is going to
be a conventional TIFF or JPEG file. (It would be possible for Fuji
cameras to store raw sensor data in a RAW format file, and have the
resampling done later).
The Fuji approach doesn't give any more information overall than a
square sensor. It *does* give more horizontal and vertical detail,
at the cost of less diagonal detail. It's the process of converting
the rotated raster to a conventional horizontal/vertical raster that
requires the image enlargement to avoid losing detail.
When the process is all done, the 6 Mpixel image produced by a Fuji
camera should have about *the same* horizontal and vertical detail
as an image captured by a full 6 Mpixel conventional sensor. That's
why the image file has to be that large. But it has only *half*
the diagonal resolution as a conventional 6 Mpixel sensor, and less
than a conventional 3 Mpixel sensor.
Dave
Dave Martindale
>> Many digital movie cameras have three CCD's, one for each
>> primary color and do not perform interpolation. Of course,
>> they mostly have relatively low resolution.
>Thanks. I didn't know that. But still cameras?
There are still cameras designed for motionless subjects that take
three successive images using a colour filter wheel and a monochrome
CCD. There are still cameras that use a mechanically-scanned linear CCD
array, like a flatbed scanner. Both of these methods produce individual
RGB samples at every pixel, but can't take instantaneous exposures.
I don't know of any still cameras that use 3 CCDs and a beamsplitting prism,
like good video cameras. The tradeoffs are different. In video, the
resolution is fixed and the CCDs are small. A prism beamsplitter is small
enough to be practical, and 3 CCDs gets you better colour reproduction
and better light sensitivity (all the incoming light is routed to one of the
CCDs, rather than having 2/3 of it absorbed by a filter), so 3 CCD cameras
produce better images at the fixed resolution. With still cameras, given a
certain pot of money, you're better off spending it on a larger single CCD
with mosaic filter than on 3 smaller CCDs and beamsplitter.
Dave
David Chien
this:
1) The lower end models, the 602 & metal-vertical cased 60 models use a
3MP SuperCCD upsampled to 6MP in the camera. Thus, effective resolution
is like a 3-4MP digicam, not a 6MP.
2) However, their new S2 SLR does in fact use a TRUE 6MP SuperCCD
sensor, upsampled to 12MP in camera. Thus, effective resolution is like
a 6-10MP digicam.
yep, looks like Fuji now has the baddest SLR out there right now with
6MP of raw sensor resolution -- naturally, we'll have to see how tests
pan out.
Tom Holub
Todd Walker <twalk...@hotmail.com> wrote:
)In article <Y3e68.279$1F2....@sea-read.news.verio.net>, do...@inl.org
)>
)> It is not a lie. The Fuji sensor is hexagonal, and to represent the output
)> of a hexagonal array on a rectangular output medium (like a computer
)> screen), you need to use double the pixels. The amount of data is
)> nearly equivalent to a rectangular sensor of the same size; it is not
)> a 6-megapixel camera (and Fuji doesn't claim it is).
)> -Tom
)>
)
)So you are saying that because of the inferior design of their CCD,
)which forces the camera to make up information that is not recorded by
)the chip, they should be able to get away with calling their cameras
)6mp? I don't think so. It's really pretty simple -- if you capture 3.3mp
)worth of information, you have a 3.3mp camera. Rather than playing
)semantics and trying to find a loophole that they can exploit, Fuji
)needs to be honest with the consumer.
No, I'm saying that because of the superior design of their CCD, which
allows the camera to capture marginally more information with the same
number of sensors (not double the resolution, but measurably more than
a rectangular sensor), they must generate a file which is double the
size. There's nothing dishonest about their approach--they say the number
of pixels on the CCD right up front.
-Tom
Guido Vollbeding
>
> I don't know of any still cameras that use 3 CCDs and a beamsplitting prism,
> like good video cameras.
The Foveon II camera does use 3 CMOS (better than CCD) and a beamsplitting prism:
http://www.foveon.net/prod_new.html
> With still cameras, given a
> certain pot of money, you're better off spending it on a larger single CCD
> with mosaic filter than on 3 smaller CCDs and beamsplitter.
YMMV (Your Mileage May Vary). I'm still waiting for a practical 3-Chip
digital camera *without* any need for color interpolation. It's the only
way to get clear, crisp, sharp and natural images. Contemporary 1-Chip
devices with mosaic filter always suffer from interpolation artifacts
and loss of detail, regardless of tricks and image postprocessing
being used. You can't reproduce the missing information with such
design, and Fuji's SuperCCD is no solution.
In professional and broadcast video production *solely* 3-Chip devices
are being used, there's no 1-Chip camera, with good reason.
Regards,
Guido
Uncle Frank
link to this thread and some representative responses to the Digital
Photography forum on Silicon Investor.
http://www.siliconinvestor.com/stocktalk/subject.gsp?subjectid=52350
The rec.photo.digital newsgroup is an amazing resource for all of us
who are interested in this great hobby. Thanks again for sharing your
insights.
uf
Bob Dietz
> Fujifilm announced 3 new cameras today, and there was a consistent
> theme; they claimed an effective resolution of about twice the pixel
> count of their ccd. As an example, from their S602Z announcement,
>
> "New 3rd Generation Super CCD sensor and advanced LSI algorithm (3.1
> million effective pixels) producing class-leading 6.03 million
> recorded pixels (2,832 x 2,128)".
>
> Is "recorded pixels" a meaningful measure, or slight of hand. I get
> the impression they may be doing something like onboard cubic
> interpolation to increase pixel count, but you could do the same
> thing with any picture file with Photoshop.
I have no idea what "advanced LSI algorithm" means.
Super CCD means that the filtered sensor array looks something like -
http://www.photoloft.com/view/Image.asp?u=1619889&s=jasc&frames=&nonav=&is=&a=1178653&i=8910669&ViewerSize=3&image6.x=89&image6.y=9
as opposed to a typical array with a bayer filter which looks something like -
http://static.howstuffworks.com/gif/digital-camera-bayer.jpg
Notice in the typical array the distribution of red, green and blue is not equal.
Notice in the hexagonal array the distribution of red, green and blue is equal.
Given the same number of sensors, it seems to me there is going to be more
usable data from the hexagonal array. Though, I doubt it's going to translate
out to twice as much real data.
--
Bob Dietz
"It needs fanatical faith to rationalize our cowardice." - Eric Hoffer
Jens Kristoffer Nielsen
wrote:
>In article <Y3e68.279$1F2....@sea-read.news.verio.net>, do...@inl.org
>says...
>
>> It is not a lie. The Fuji sensor is hexagonal,
>
>No it's not. It's a rectangular array rotated by 45 degrees.
A bit more complicated, I'm afraid. It is a a roughly rectangular
array of hexagonal sensors with the different lines staggered in a
honeycomb pattern. They claim that it allows them to maximize the area
of each sensor, thus increasing the surface capable of detecting
light. It still doesn't record twice the amount of data, though. It is
correct that if you don't double the amount of pixels in the finished
file, you actually throw away data. But it is not really double the
amount of information. Thus you really cannot compare the number of
megapixels between a SuperCCD and a CCD. It might be that a pixel in
the file from the SuperCCD is worth somewhere between 0.55 and 0.75
pixel from a normal CCD - but that is just a guesstimate.
/Kristoffer
Peter A. Stavrakoglou
interpolate with software the way the Fuji does with the hardware
interpolation.
"Tom Holub" <do...@inl.org> wrote in message
news:6Ig68.312$1F2....@sea-read.news.verio.net...
John Smitty
piece of marketing. Fuji's current cameras are not helped by this
iterpolation and it won't be any better this bunch. Go to the
comparometer http://www.imaging-resource.com/ and compare the 6900
images to a good 3 or 4 mp camera like the G2. The 6900 colors are bad
with a red cast and the detail is not nearly as good. These new cameras
are just a rehash of the old technology. You can add fancy features all
you want. If the image quality isn't there it doesn't matter. You can
interpolate all you want. Give me a real 4 or 5 mp camera, not this
hocus pocus junk. These cameras will be a disappointment, even more
than the 5000. Yopu heard it here first. Fuji is second tier and will
continue to be until they get rid of this bogus super ccd. Fuji fans
will try to explain why every flaw is ok and can be fixed in PS and try
to justify the camera just like the nikon crowd does with the 5000.
The new Fuji cameras are Garbage. They will flop. Be sure to come back
and tell me I'm right when these cameras are released- well past the
current scheduled date. Thats right there will be delays, you heard it
hear first. Can't wait to hear how right I am....
Peter A. Stavrakoglou
well. And many consumers are "voting" with their wallets - it is a wildly
popular camera that is difficult to find in stock, at least in the US.
"Todd Walker" <twalk...@hotmail.com> wrote in message
news:MPG.16c32ed27...@news-server.jam.rr.com...
Walt
<ne...@childsupportanalysis.co.uk> wrote:
>While each of them merits analysis, it is wise to keep them separate. Apart from
>anything else, while you can resize in Photoshop, possibly to better effect, as
>far as a I know you can't do colour interpolation except in the camera.
No, you can definitely do the color interpolation outside the camera, at least
if the camera saves raw files the way the Olympus and Canon do. The product I'm
working on will be offering multiple interpolation methods as part of the
camera loading mechanism, too.
Walt
Walt
wrote:
>In professional and broadcast video production *solely* 3-Chip devices
>are being used, there's no 1-Chip camera, with good reason.
Yeah... but you have to be honest about that good reason: It's because
broadcast video has such terrible resolution to start with, that they can't
afford to drop 2/3rds the data, especially vertically. :)
NTSC: No
Thanks to
Sharp
Color
walt
Mark Grebner
> much better results than the naive claim of 1/3 the actual number of
What naive claim is that? Personally, I claim that the monochrome
linear resolution is roughly equal to the pixel count divided by
sqrt(2). That is, a camera that claims to be "1200 x 1600" can
actually resolve about 850 lines (or 425 line pairs) in the narrow
dimension.
By "resolve", I mean that if you try to count the lines, you get the
correct number, even though you'll have to put up with severe aliasing
at the 425-pair limit.
Maybe the "1/3" claim is for each color?
Guido Vollbeding
>
> Yeah... but you have to be honest about that good reason: It's because
> broadcast video has such terrible resolution to start with, that they
> can't afford to drop 2/3rds the data, especially vertically. :)
Right, but why should *we* afford to drop 2/3rds of data?
I think the multi-megapixel boost is the wrong way to go.
We also need the 3-Chip advantage, and someday users will
notice and demand it (when they realize that more megapixels
do not help).
> NTSC: No
> Thanks to
> Sharp
> Color
I thought it was "Never The Same Color" :).
Regards,
Guido
David Oddie
<nto...@optonline.net> wrote:
>Many reviews of this camera indicate that the hardware interpolation works
>well.
Not well enough to call it six mega-pixels though.
If you read the reviews the ones I have read say the resolution is
higher than other 3.3 mp camera but is not up to 5 mp cameras, being
somewhere in the 4 mp range.
A good analogy is AMD with their current top-line CPU's.
Their model numbers are, for example, 1600 or 2000 and are supposed to
indicate that their CPU's equal to an Intel Pentium 4 running at that
speed i.e. 1.6 Mhz or 2.0 Mhz.
The independent tests of these CPU's show their claims are correct, so
IMO it is valid for them to stick numbers on them that indicate they
are equal to certain P4 CPU's.
The same cannot be said for Fuji and its super CCD.
The tests show it is clearly does not possess the resolving power the
current 5 mp cameras (let alone the true 6 mp camera about to appear)
so for them to say, in any way, the camera is a 6 mp camera is wrong.
If they said its 3.3 with an effective measurable resolution of 4 mp
(or whatever it is) then they would be playing the game and would not
get much criticism.
If, as stated by one poster in this thread that the design of the
super CCD requires a file that "contains" six mega-pixels despite the
fact its true resolving power is less than that, what is the point?
You have got a camera of around 4 mp true resolution but have to
manipulate larger files than from a true 4mp camera. Lower resolution
(than implied) + larger file = inferior product to true 4 mp cameras
IMO.
Yes I am sure the colours are good an there are happy owners but if
you need 6 mp resolution the only place you will find it is with a
Contax N1 or Fuji's own S2 Pro - without interpolation active :-)
Dave
--
It may be that your sole purpose in life is simply to serve as a
warning to others.
Remove the uppercase N O S P A M to reply via email.
AM
nospam.spa...@jkn.dk says...
> A bit more complicated, I'm afraid. It is a a roughly rectangular
> array of hexagonal sensors
Even that is not accurate - the sensors are octagonal, not hexagonal.
> with the different lines staggered in a
> honeycomb pattern. They claim that it allows them to maximize the area
> of each sensor, thus increasing the surface capable of detecting
> light. It still doesn't record twice the amount of data, though. It is
> correct that if you don't double the amount of pixels in the finished
> file, you actually throw away data. But it is not really double the
> amount of information. Thus you really cannot compare the number of
> megapixels between a SuperCCD and a CCD. It might be that a pixel in
> the file from the SuperCCD is worth somewhere between 0.55 and 0.75
> pixel from a normal CCD - but that is just a guesstimate.
It is true that the higher the fill ratio (ratio of light sensitive area
to total pixel area) the more sensitive the camera. So if Fuji manages to
increase the fill ratio, that's good, but then you could find a way to
increase the fill ratio also in a normal CCD.
Fuji claims that there is a gain due to the different orientation of the
CCD axes, because (they say) humans are more sensitive to certain
directions, and could be an advantage if it's true.
Tom Holub
David Oddie <Dave...@mNaOiSlPaAnMdnews.com> wrote:
)On Fri, 01 Feb 2002 02:22:27 GMT, "Peter A. Stavrakoglou"
)<nto...@optonline.net> wrote:
)
)>Many reviews of this camera indicate that the hardware interpolation works
)>well.
)
)Not well enough to call it six mega-pixels though.
They don't claim it's a 6 megapixel camera. They say right in the
first line that it has 3.1 million effective pixels. It generates a
6-megapixel image because it HAS to due to the nature of the hexagonal
CCD.
)If, as stated by one poster in this thread that the design of the
)super CCD requires a file that "contains" six mega-pixels despite the
)fact its true resolving power is less than that, what is the point?
There are two points:
1) Fuji's SuperCCD cameras tend to deliver more measurable resolution
than comparable rectangular-array cameras. (Again, not double, but
more).
2) Fuji's SuperCCD cameras deliver better color than comparable
rectangular-array cameras.
Given these two points, it seems reasonable to assert that Fuji's
SuperCCD design has noticable benefits over rectangular-array designs.
)Yes I am sure the colours are good an there are happy owners but if
)you need 6 mp resolution the only place you will find it is with a
)Contax N1 or Fuji's own S2 Pro - without interpolation active :-)
I have not used the S2 Pro, but if it's anything like the 4900/6900,
you will lose if you don't use the highest resolution. On my 4900,
if I choose 1600x1200 resolution (just under the pixel count of the CCD),
pictures have significantly less detail than in 2400x1800 mode.
Naked-eye-on-print noticable.
-Tom
Todd Walker
says...
> In article <u97l5uc3plf9v31pi...@4ax.com>,
> David Oddie <Dave...@mNaOiSlPaAnMdnews.com> wrote:
> )On Fri, 01 Feb 2002 02:22:27 GMT, "Peter A. Stavrakoglou"
> )<nto...@optonline.net> wrote:
> )
> )>Many reviews of this camera indicate that the hardware interpolation works
> )>well.
> )
> )Not well enough to call it six mega-pixels though.
>
> They don't claim it's a 6 megapixel camera. They say right in the
> first line that it has 3.1 million effective pixels. It generates a
> 6-megapixel image because it HAS to due to the nature of the hexagonal
> CCD.
Then why does it have a 3.1mp mode?
> )If, as stated by one poster in this thread that the design of the
> )super CCD requires a file that "contains" six mega-pixels despite the
> )fact its true resolving power is less than that, what is the point?
>
> There are two points:
>
> 1) Fuji's SuperCCD cameras tend to deliver more measurable resolution
> than comparable rectangular-array cameras. (Again, not double, but
> more).
Can you cite some sources for this assertion?
> 2) Fuji's SuperCCD cameras deliver better color than comparable
> rectangular-array cameras.
Oversaturated isn't better.
> Given these two points, it seems reasonable to assert that Fuji's
> SuperCCD design has noticable benefits over rectangular-array designs.
I respectfully disagree. This image should put an end to this
discussion:
http://makeashorterlink.com/?P2EA26C5
That is from an F601. If you think that has more resolution than a G2,
CP5000, F707, etc then I'll eat my shoe.
There are more available here:
http://myalbum.ne.jp/cgi-bin/a_menu?id=fa782151
The pictures in 3.3mp mode look decent. The ones in 6mp mode look like
crapola.
Chris Brown
Tom Holub <do...@inl.org> wrote:
>They don't claim it's a 6 megapixel camera. They say right in the
>first line that it has 3.1 million effective pixels. It generates a
>6-megapixel image because it HAS to due to the nature of the hexagonal
>CCD.
POI - SuperCCD is not hexagonal. The array is rectangular (each sensor has 4
neighbours), the sensors themselves are octagonal. There isn't a hexagon in
sight.
--
/* _ */main(int k,char**n){char*i=k&1?"+L*;99,RU[,RUo+BeKAA+BECACJ+CAACA"
/* / ` */"CD+LBCACJ*":1[n],j,l=!k,m;do for(m=*i-48,j=l?m/k:m%k;m>>7?k=1<<m+
/* | */8,!l&&puts(&l)**&l:j--;printf(" \0_/"+l));while((l^=3)||l[++i]);
/* \_,hris Brown -- All opinions expressed are probably wrong. */return 0;}
Tom Holub
Todd Walker <twalk...@hotmail.com> wrote:
)In article <llz68.369$1F2....@sea-read.news.verio.net>, do...@inl.org
)says...
)> In article <u97l5uc3plf9v31pi...@4ax.com>,
)> David Oddie <Dave...@mNaOiSlPaAnMdnews.com> wrote:
)> )On Fri, 01 Feb 2002 02:22:27 GMT, "Peter A. Stavrakoglou"
)> )<nto...@optonline.net> wrote:
)> )
)> )>Many reviews of this camera indicate that the hardware interpolation works
)> )
)> )Not well enough to call it six mega-pixels though.
)> They don't claim it's a 6 megapixel camera. They say right in the
)> first line that it has 3.1 million effective pixels. It generates a
)> 6-megapixel image because it HAS to due to the nature of the hexagonal
)> CCD.
)
)Then why does it have a 3.1mp mode?
It also has a 640x480 mode. So?
As I posted elsewhere, if you actually compare images, the so-called
"interpolated" mode has significantly, obviously more detail than the
so-called "uninterpolated" mode. At least for the 4900 and 6900.
)> )If, as stated by one poster in this thread that the design of the
)> )super CCD requires a file that "contains" six mega-pixels despite the
)> )fact its true resolving power is less than that, what is the point?
)>
)> There are two points:
)>
)> 1) Fuji's SuperCCD cameras tend to deliver more measurable resolution
)> than comparable rectangular-array cameras. (Again, not double, but
)> more).
)
)Can you cite some sources for this assertion?
The dpreview resolution analysis, for example.
)> 2) Fuji's SuperCCD cameras deliver better color than comparable
)> rectangular-array cameras.
)
)Oversaturated isn't better.
I don't find that my Fuji produces oversaturated images. Certainly a
lot of other people, and a number of camera review sites, have commented
on the high quality of color reproduction of the Fuji SuperCCD cameras.
But to some extent this is a matter of taste.
-Tom
Chris Brown
>> first line that it has 3.1 million effective pixels. It generates a
>> 6-megapixel image because it HAS to due to the nature of the hexagonal
>> CCD.
>
>Then why does it have a 3.1mp mode?
My D30 has a 1.x megapixel mode. Can't imagine why I'd ever use it, but it's
there. Fact is, pictures taken in that mode will contain less information
than ones taken in 3.e megapixel mode, and not a SuperCCD in sight.
>> 2) Fuji's SuperCCD cameras deliver better color than comparable
>> rectangular-array cameras.
>
>Oversaturated isn't better.
They do appear to gain a resolution advantage on Phil Askey's reviews. YMMV.
Disclaimer - I do not own a Fuji digital camera, and am not currently
contemplating buying one.
Todd Walker
says...
> )> 1) Fuji's SuperCCD cameras tend to deliver more measurable resolution
> )> than comparable rectangular-array cameras. (Again, not double, but
> )> more).
> )
> )Can you cite some sources for this assertion?
>
> The dpreview resolution analysis, for example.
From the DPReview review of the 6900:
"That said the 6900Z isn't really able to render any finer detail than a
good 3 megapixel digital camera. Its in-camera 'processing' is definetly
very sophisticated and can generate very pleasing six megapixel images,
however they still really only have 3 megapixels of detail."
and from the conclusion of that review:
"Resolution as good as the best 3 megapixel digital cameras (performed
well on our resolution tests)"
As good as != better.
> )> 2) Fuji's SuperCCD cameras deliver better color than comparable
> )> rectangular-array cameras.
> )
> )Oversaturated isn't better.
>
> I don't find that my Fuji produces oversaturated images. Certainly a
> lot of other people, and a number of camera review sites, have commented
> on the high quality of color reproduction of the Fuji SuperCCD cameras.
> But to some extent this is a matter of taste.
> -Tom
I'll give you that one. Phil's review does mention that it has excellent
color reproduction but also notes the lack of saturation control:
"Excellent colour reproduction, probably the best in its class (caveat:
some saturation control would be nice)"
I'm not trying to knock Fuji Tom. I actually considered the 6900 when I
bought my Pro90. The Canon had the feature set that I wanted so I bought
it. I just don't like Fuji attempting to mislead people regarding the
resolution of their cameras. That plus I like a healthy debate ;-) No
harm, no foul...
Tom Holub
Todd Walker <twalk...@hotmail.com> wrote:
)In article <BEA68.374$1F2....@sea-read.news.verio.net>, do...@inl.org
)says...
)> )> 1) Fuji's SuperCCD cameras tend to deliver more measurable resolution
)> )> than comparable rectangular-array cameras. (Again, not double, but
)> )> more).
)> )
)> )Can you cite some sources for this assertion?
)> The dpreview resolution analysis, for example.
)
)From the DPReview review of the 6900:
)
)"That said the 6900Z isn't really able to render any finer detail than a
)good 3 megapixel digital camera. Its in-camera 'processing' is definetly
)very sophisticated and can generate very pleasing six megapixel images,
)however they still really only have 3 megapixels of detail."
)
)and from the conclusion of that review:
)
)"Resolution as good as the best 3 megapixel digital cameras (performed
)well on our resolution tests)"
)
)As good as != better.
The comments you quote are referring to the performance on a series of
test shots. The 6900's pictures obviously show finer detail than the
Nikon 995 he's comparing to, but he attributes the improved resolution
to better lens performance on the 6900. (And indeed it is difficult
to separate lens performance from CCD performance).
If you specifically look at the resolution measurement page
<http://www.dpreview.com/reviews/fuji6900z/page14.asp> you'll note
that the measured resolution for the 6900 is higher than for the other
3-megapixel cameras tested, and is nearly comparable to the
4-megapixel Sony DSC-S85.
He also specifically notes that a 6-megapixel shot downsampled to 3
megapixels has more detail than a shot taken in 3-megapixel mode.
)I'm not trying to knock Fuji Tom. I actually considered the 6900 when I
)bought my Pro90. The Canon had the feature set that I wanted so I bought
)it. I just don't like Fuji attempting to mislead people regarding the
)resolution of their cameras.
I don't think the Fuji is the right camera for everyone. But I don't
think Fuji is misleading anyone. The images the camera produces are
double the pixel count of the CCD due to requirements of the
technology. I don't think Fuji's camera line would be better if it
used a more traditional CCD. And Fuji advertises the size of the CCD
quite prominently. You need to "think outside the rectangle."
-Tom
Mark Herring
Search on Google groups for some previous exchanges on this subject.
Bottom line: you dont add information by increasing the file size.
The one reasonable argument I have heard is that the tilted array
increases horiz and vertical resoltuion at the expense of diagonal.
IMHO, that does nothing for the appearance of a random scene from
nature.
I still regard the Fuji claims as little more thatn marketing
hype---it would be good to see some test results.
-Mark
n 30 Jan 2002 23:13:01 -0800, uncle...@hotmail.com (Uncle Frank)
wrote:
>Fujifilm announced 3 new cameras today, and there was a consistent
>theme; they claimed an effective resolution of about twice the pixel
>count of their ccd. As an example, from their S602Z announcement,
>
>"New 3rd Generation Super CCD sensor and advanced LSI algorithm (3.1
>million effective pixels) producing class-leading 6.03 million
>recorded pixels (2,832 x 2,128)".
>
>Is "recorded pixels" a meaningful measure, or slight of hand. I get
>the impression they may be doing something like onboard cubic
>interpolation to increase pixel count, but you could do the same
>thing with any picture file with Photoshop.
>
>uf
David Oddie
>They don't claim it's a 6 megapixel camera. They say right in the
>first line that it has 3.1 million effective pixels. It generates a
>6-megapixel image because it HAS to due to the nature of the hexagonal
>CCD.
It does not generate a 6 mp image.
It actually generates a 3.1 mp image and manipulates it to yield a
larger file which contains an image that is of seemingly higher
resolution than 3.1 mp but is not 6 mp of resolution. It has six mp
in the file but not six mp of resolution.
So what that file contains is another matter. By using the 6 mp
figure, Fuji clearly want you to believe the file contains and image
that is equivalent to an image that would arise from a genuine 6 mp
CCD.
The reason I say this is the fact 6 mp is mentioned at all gives the
impression it is a 6 mp camera. It does not matter how they say what
they say - the mere fact 6 mp is mentioned is enough to confuse.
If, because the super CCD works a certain way, it generates a larger
file - fair enough, but file sizes are not measured in mega-pixels but
in bytes. So your 3.1 mega-pixel camera will, in a certain mode,
generate a larger file than any other 3.1 mp camera.
If they said in the "highest quality" mode file sizes would be around
X megabytes compared to when shooting as a 3.1 mp camera they would be
Y size and smaller, fair enough.
But they don't - they bring up a unit of measure that is used to
measure digital image resolution and not file size.
They want people to belive that because the file is big enough to hold
a six mega-pixel image that they are getting the resolution of a
six-mega pixel camera.
Dave
--
Remove the upperase N O S P A M letters from the email address
to reply via email
Walt
wrote:
>I think the multi-megapixel boost is the wrong way to go.
>We also need the 3-Chip advantage, and someday users will
>notice and demand it (when they realize that more megapixels
>do not help).
I'm *completely* with you on this one. :)
>> NTSC: No
>> Thanks to
>> Sharp
>> Color
>
>I thought it was "Never The Same Color" :).
...and
NTSC: Never
Twice the
Same
Color
...and
NTSC: Needs
Tweaking in
Saturation and
Chroma
...and probably many others. I'm of a mind that you can't really learn to hate
NTSC properly until you've tried to make it as good as it can be, and you find
out that everything, but *everything*, works against you.
Walt
Ian Burley
length shift problem. Canon's large 4MP sensor is (in the 1D), to me, more
valuable.
I'd also like to see the noise characteristics of this new 6MP SuperCCD -
I'll be the first to congratulate Fuji if it's as good as Canon's 3MP and
4MP SLT chips.
Ian
--
Digital Photography Now
UK-based digital photo Web magazine
www.dp-now.com
"David Chien" <chi...@uci.edu> wrote in message
news:3C599159...@uci.edu...
Kennedy McEwen
<will...@adobe.com> writes
>
>DPReview reported resolution figures for the 3MP S1: 1300/1200/1000
>horizontal / vertical / diagonal lines vs. 900/900/900 for the 3MP Nikon 990
>and 1000/1150/1000 for the D30. This is consistent with "not 2X but not
>a complete fabrication".
>
Actually Russell, that is, more or less, exactly equivalent to 2X!
Vertical and horizontal resolution is about 1.4x in each axis, comparing
with the Nikon and a little short of that in the D30. So you need 1.4x
as many pixels in the vertical AND horizontal, giving a total of 2X.
The diagonal resolution is simply a consequence of the H & V resolution
and the structure of the device.
It is a common misconception that a 2Mpix camera has twice the
resolution of a 1Mpix camera when, in fact, it is only 1.4X in each
axis.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)
Kennedy McEwen
<alfred...@yahooREMOVE.com> writes
>In article <3C595D2C...@usfamily.net>, stau...@usfamily.net
>says...
>> What really bugs me is that there is an ISO standard for measuring
>> resolution from a digital camera. No one in the consumer digicam
>> industry that I know uses it. We, the user community, need to push
>> manufacturers to use this procedure. Mags can help. Whenever they do a
>> review on a new camera, they should press manufacturers for ISO res
>> data. They should not publish ANYTHING about 'resolution' unless the
>> ISO test results are provided. Or, the mags could set up to do the test
>> themselves.
>
>What is this ISO standard ?
The standards for measuring the resolution of both scanners and digital
cameras is set out in ISO-16067 and ISO-12233 and is based on a method
which I have been using for resolution measurements on digital imaging
sensors for over 10 years. The technique was originally developed by Dr
K. Murphy of the Royal Signals and Radar Establishment (now Qinetiq),
Malvern, UK in the mid-1980s, although none of the references in the ISO
standard acknowledge this. I have a copy of Mr Murphy's original
publication which predates any of the references provided in the ISO
document, however I was not even aware that ISO were contemplating such
a standard, or considering this methodology until it was brought to my
attention about 6 months ago - long after the deadlines for reference
citations closed.
The technique concerns the use of a single sharp edge, slightly slanted
relative to the vertical and horizontal axes, which transitions from
near full black to near full white. Imaging this edge with a camera
produces an Edge Spread Function, a measure of how much the camera
resolves or blurs the edge. Spatial differentiation of the ESF results
in a Line Spread Function, a measure of how the camera resolves and
blurs lines. By orienting the spatial differentiation along the axis
closest to the edge rather than the axis across it, the ESF is
essentially oversampled by the inverse of the tangent of the inclination
angle. This enables the LSF to be determined at many times the limiting
resolution of the digital structure, effectively multiplying the Nyquist
limit. Using a standard mathematical process known as Fourier Analysis,
the Modulation Transfer Function is determined directly from the LSF.
MTF is the standard measure of resolution for all optical components,
including lenses, film, CCDs and complete systems. Since the LSF has
been oversampled, the MTF is determined without the significant effects
of aliasing which plague alternative methods of assessing resolution,
particularly close to the resolution limits of the imaging sensor.
Imaging conventional test patterns, such as line pair patterns, is
particularly prone to error due to phase dependent sampling effects,
making it virtually worthless as a resolution measurement tool on
digital imaging sensors. To avoid confusion with standard methods of
obtaining MTF from analogue components such as optical lenses etc., the
ISO standard refers to the measured parameter as the Spatial Frequency
Response, or SFR, but notes that it is simply the digital equivalent of
MTF.
The ISO standard includes the recommended test image and the ISO web
site has digital versions of this and the analysis software, so anyone
can test their digital camera and scanner using these files, which are
freely downloadable from the net - or at least they were when I looked a
few months ago.
On the main subject of this thread, whether the Fuji CCD produces higher
resolution images than other equivalent chips, this is well documented
and has been proven on many occasions by independent authorities. For
that reason Fuji's SuperCCD won the year 2000 IEEE Walter Kosonocky
Technical Achievement Award. For reference, Walt Kosonocky was one of
the pioneers of imaging CCDs, working at RCA, Rome Air Defence and
ultimately at the Sarnoff Labs in New Jersey. Neither he, who I had the
pleasure of meeting and working with on several occasions, nor the IEEE
are known for piling praise on technologies which do not merit it!
The Fuji CCD does not actually resolve more information than alternative
designs, although its cell structure does permit a higher packing
density than is possible with a conventional x-y matrix chip, resulting
in higher signal to noise ratios and better colour reproduction. The
key to the increased resolution of Fuji images lies in optimising the
CCD's high resolution axes to real world images, and it does this at the
expense of throwing out the conventional x-y pixel orientation. If you
want to stick to a conventional display orientation, as most of us do
for obvious compatibility reasons, then it should be no surprise that
you need more of these sub-optimal resources to support the full
performance of the optimised image. The fact that you need twice as
many is simple geometry.
It might come as a surprise to some on this thread, but in the real
world has a dominance of high resolution aligned to the horizontal and
vertical axes.
"Sir Isaac Newton told us why,
An apple falls down from the sky,
And from this fact its very plain,
All other objects do the same.
A brick, a bolt, a bar, a cup,
Invariably fall down, not up..." or at 45deg!
The world you see is dominated by horizontal (hence the very term
horizon-tal!) and vertical detail as nature and man use the power of
gravity to make their structures - but there is much less detail at
other angles. This means that a majority of natural images have their
finest details aligned to the vertical and horizontal axes. This has
nothing whatsoever to do with human vision or perspectives - it is real,
and a direct consequence of gravity.
If you Fourier Transform an image you can see where the finest detail
lies - almost invariably the detail is clustered along the vertical and
horizontal axes. Fuji used to have an example image which demonstrated
this on their brochure, which was available as a pdf file from their web
site - sccd.pdf - though I do not know if it is still there. I do know
that I have analysed thousands of real images prior to the Fuji
development which independently supports this. Certainly, there are
cases where this is not true - a photograph of a surface of foaming soap
bubbles has the detail distributed randomly around all directions
without any preference but, in most normal images (even in simple
portraiture), there is a strong dominance in the horizontal and vertical
axes of the frame.
In contrast to real world images, x-y matrix systems (both image chips
and displays) have their highest resolution at 45deg to the horizontal
and vertical axes. If you measure this on a CCD imager, and I have on
numerous occasions, the resolution at 45deg can be as much as 1.414x
higher than the resolution in the vertical or horizontal axes - lens
resolution permitting.
You just need to count the number of discrete lines in, say 1cm, you can
draw through an x-y matrix of dots to demonstrate part of this to
yourself, but there are other issues involved which are less easily
demonstrated. Alternatively, take your digital camera and photograph a
line test pattern - a series of long lines printed on a page that are
just as thick as the gaps between them. Set the lines vertical or
horizontal and move back from them until they are just unresolved or
beginning to be aliased by your digital camera. Assuming that your CCD,
not your camera lens, is the limiting factor (which may not be the case
for some of the small commercial 3Mpix cameras) simply rotate the camera
through 45deg and, as if by magic, the lines are now resolved, alias
free. That is the principle of one aspect of the SuperCCD - throwing
less information away than the conventional arrangement forces you to
do.
So a conventional CCD actually captures a real world image in exactly
the worst possible orientation - its lowest resolution is aligned with
the finest detail in the typical image, whilst its highest resolution is
aligned with the lowest concentration of detail. All of this has been
known about for years and it has long been recognised that if a CCD
could be manufactured on a 45deg axis then it would provide better
resolution in most cases. Fuji were the first company to take these
ideas forward to a real commercial product and demonstrate that a
decades old theory could actually be implemented in practice and the
benefits realised. That is why they won the IEEE Kosonocky Technical
Achievement Award - lots of people had speculated about it, but none had
actually gone out and done it. In addition, at the same time, Fuji
developed octagonal CCD cells (not hexagonal as some in this thread have
referred to them - there are 8 sides). These also have resolution
advantages, since the octagonal structure, whilst permitting higher
packing density of the cells, also has a spatial resolution
characteristic which well matches the circular symmetry of the lens, and
its spatial characteristics.
The SuperCCD has a real resolution advantage in the axes which dominate
images of approximately 1.414x in each axis. The lens may reduce this
advantage depending on the particular camera that the SuperCCD is fitted
to, since lenses are not infinite resolution components. Assuming the
lens is not the limiting factor, to correctly display this image without
loss of resolution in the horizontal and vertical axes you need at least
1.414x as many pixels in each axis - not surprisingly that equates to a
total number of pixels which is twice the number of cells in the
SuperCCD. Hence a 3Mpix SuperCCD requires 6Mpixels to display it, if
the benefit of the SuperCCD is not to be discarded.
This isn't the same as saying the image is EQUIVALENT to that obtained
by a conventional 6Mpix camera - for one part, a 6Mpix camera has almost
twice the 45deg diagonal resolution of the 3Mpix SuperCCD and, whilst
this is not usually the dominant axis in an image, this is additional
information adding to image quality. The 3Mp SuperCCD has virtually the
same vertical and horizontal resolution as a 6Mpix conventional CCD, but
whilst the resolution of the SuperCCD falls to its minimum as it rotates
to 45deg, the conventional 6Mpix chip resolution increases to its
maximum. The difference is that the conventional approach has its worst
performance where it is really needs its best.
There is nothing magic or deceptive about the Fuji approach, and it
certainly ISN'T interpolation. It is a simple consequence of attempting
to display an optimal image format (optimised to real world conditions)
on a sub-optimal, but conventional, display format.
If you had the ability to move the pixels on your display into a 45deg
axis then you could get the same horizontal and vertical resolution with
just the number of pixels as cells on the CCD, since display pixels have
the same resolution properties as CCD cells - the highest resolution
they can support is at 45deg to their axes. *IF* you want to retain the
resolution advantage on a conventional display that the SuperCCD
provides then you need to display the image in twice as many pixels.
However it does not need to be stored as twice the file size except for
conventional file formats. The image can be stored in raw format, with
only the real number of pixels used, but require converting to x-y
format for computer storage, since all conventional file formats use the
less efficient (in resolution terms) x-y structure.
You can look at the Fuji approach in one of two ways:
* the SuperCCD is a better match to both real world images and camera
characteristics than conventional CCDs
or
* the SuperCCD approach eliminates many of the resolution losses which
we accept as 'normal' in conventional camera designs. Fuji have
demonstrated that we don't need to accept these losses - it doesn't have
to be this way!
Also note that the SuperCCD advantage is scaleable - just because 5 and
6Mpixel cameras are coming along in conventional formats doesn't mean
that the SuperCCD is a thing of the past. On the contrary, the higher
packing density of the SuperCCD means that it is easier for Fuji to
develop 5 or 6Mpix versions of this in due course - producing 10 and
12Mpix images respectively.
Disclaimer: I do not own, though I have tested and might consider
buying, a Fuji SuperCCD camera and have no connection with Fuji in any
way. I design digital imaging sensors for, amongst other things,
military surveillance systems and, clearly, optimising the components of
the system to work together is critical in such designs.
Rick Decker
undestand from one reading but someone should post this on Canon/Nikon/FUji
fourms. I tired but it ws too long. Very UNbiased although you know all you
english are extremely biased towards the Fuji S1 (it seems they are mostly
wedding photographers). My limited experience with the S1 has been very
positive so far.
Barry Pearson
news:k8vj5u839gi5qnoqr...@4ax.com...
I didn't realise that RAW from any camera provided access to the
pre-colour-interpolated state. (But I hadn't studied it - it isn't something
that the camera manufacturers promote a lot).
Will your product be a Photoshop plug-in or something else? (I'm not clear what
you mean by "part of the camera loading mechanism").
--
Barry Pearson
http://www.barrypearson.co.uk/photography/
Barry Pearson
[snip]
> NTSC: No
> Thanks to
> Sharp
> Color
I was always told is stood for "Never Twice the Same Colour".
But I was never sure how much that was simply propaganda from the PAL people.
Chris Brown
Barry Pearson <ne...@childsupportanalysis.co.uk> wrote:
>
>I didn't realise that RAW from any camera provided access to the
>pre-colour-interpolated state. (But I hadn't studied it - it isn't something
>that the camera manufacturers promote a lot).
Depends on the camera manufacturer. Canon make quite a big deal about it. I
wouldn't really describe post-intepolated data as raw.
Barry Pearson
news:feam3a...@gecko.localdomain...
> In article <QKx78.9241$BU3.1...@news6-win.server.ntlworld.com>,
> Barry Pearson <ne...@childsupportanalysis.co.uk> wrote:
> >
> >I didn't realise that RAW from any camera provided access to the
> >pre-colour-interpolated state. (But I hadn't studied it - it isn't
> >something that the camera manufacturers promote a lot).
>
> Depends on the camera manufacturer. Canon make quite a big deal
> about it. I wouldn't really describe post-intepolated data as raw.
Ah! I just found:
http://www.colorshots.com/cs101e/html/tipps_raw.html
Clever - by collecting pre-colour-interpolated data they get smaller file sizes.
(It just isn't something that had come to my attention. While manufacturers talk
about the advantages of raw output, I hadn't realised that this meant
pre-colour-interpolation).
Walt
<ne...@childsupportanalysis.co.uk> wrote:
>Will your product be a Photoshop plug-in or something else?
No, WinImages is a stand-alone application the same way Photoshop is, and the
Oly and Canon RAW loaders are dedicated parts of WinImages. You can check
WinImages out here:
http://www.blackbeltsystems.com/
> (I'm not clear what you mean by "part of the camera loading mechanism").
I just meant that when you load an image in RAW format, you'll have an
interesting range of choices as to what kind of interpolation you want to use.
There are many color and multiple monochrome interpolation methods; our goal is
to give you a bunch of choices so you can pick one that works best with a
particular image.
Walt
Arjen van Andel
they don't want us to know? ;o)
Anyway, here are some links:
http://home.fujifilm.com/products/digital
http://home.fujifilm.com/products/digital/sccd/faq.html
http://home.fujifilm.com/products/digital/sccd/index.html sccd.pdf (4MB).
--
Arjen.
Justin
> The world you see is dominated by horizontal (hence the very term
> horizon-tal!) and vertical detail as nature and man use the power of
> gravity to make their structures - but there is much less detail at
> other angles. This means that a majority of natural images have their
> finest details aligned to the vertical and horizontal axes. This has
> nothing whatsoever to do with human vision or perspectives - it is real,
> and a direct consequence of gravity.
Fuji are partly true when they say that it's to do with human vision.
I can't remember which for sure. But I think horizontal structures stand out
stronger in the human vision. I learnt it in design. It probably ties into
the above. Most detail is probably more in the horizontal, than the
vertical. So the brain/eye's have evolved to perceive it stronger.
Justin.
Kennedy McEwen
<jb...@soupisgoodfood.nospam.net> writes
response in human vision to horizontal and vertical edges, though I seen
no evidence of it, however the effect which Fuji are exploiting has
nothing whatsoever to do with human vision - just physics.
There is no reason at all why their camera would be better just by
emulating one particular mammal's vision system - after all, if it only
produced a sharp image over the central 1mm of the CCD would you think
that an improvement? Its exactly what the human eye does - most of the
image formed on the retina is dominated by the aberrations of the simple
lens in the human eye. Nature isn't always better!
Rick Decker
time. Obviously by an expert in the business who really knows their
stuff. We should have more posts of this quality. Terms such as "hocus
pocus", "not nearly as good", "image quality isn't there", "a real 4 or 5
mp camera", "can't wait to hear how right I am", and many others enhance
this person's reputation as a true expert in the industry.
John Smitty wrote:
> These new releases do not exist yet everyone is excited by a published
> piece of marketing. Fuji's current cameras are not helped by this
> iterpolation and it won't be any better this bunch. Go to the
> comparometer http://www.imaging-resource.com/ and compare the 6900
> images to a good 3 or 4 mp camera like the G2. The 6900 colors are bad
> with a red cast and the detail is not nearly as good. These new cameras
> are just a rehash of the old technology. You can add fancy features all
> you want. If the image quality isn't there it doesn't matter. You can
> interpolate all you want. Give me a real 4 or 5 mp camera, not this
> hocus pocus junk. These cameras will be a disappointment, even more
> than the 5000. Yopu heard it here first. Fuji is second tier and will
> continue to be until they get rid of this bogus super ccd. Fuji fans
> will try to explain why every flaw is ok and can be fixed in PS and try
> to justify the camera just like the nikon crowd does with the 5000.
>
> The new Fuji cameras are Garbage. They will flop. Be sure to come back
> and tell me I'm right when these cameras are released- well past the
> current scheduled date. Thats right there will be delays, you heard it
> hear first. Can't wait to hear how right I am....
Mark Herring
Yes, john is a bit colorful in the language, but the truth is there.
Super CCD and interpolation are marketing gimmicksOn Wed, 06 Feb 2002
Woody Windischman
That much is true, but we shouldn't forget that the S2 has over 6 million
"real" pixels, and should still be a good performer if they've taken care of
the saturation issues. If someone is impressed by the 12 million FujiPixels
even after all of this discussion, though, that's their problem. Let them
see what the difference is between in-camera and PhotoShop interpolation.
- Woody -
"Mark Herring" <mherr...@sprynet.com> wrote in message
news:mbl16ucbm6s7eacoq...@4ax.com...
Kennedy McEwen
<mherr...@sprynet.com> writes
>
>Yes, john is a bit colorful in the language, but the truth is there.
>Super CCD and interpolation are marketing gimmicks
Azimov's first law of science fiction:
"Any technology sufficiently advanced will be considered magic by those
too primitive to understand it."
from which we derive Herring's postulate:
"Any technology sufficiently advanced that Mark doesn't understand it
must be a marketing gimmick, whether it actually produces measurable
benefits in independent tests or not!"
Perhaps you would like to explain why the SuperCCD was awarded top
position for TECHNICAL merit by the IEEE 18 months ago, a group of
people who exhaled more knowledge about imaging technology in their
morning jog than you are ever likely to learn in your entire life and
who have NOTHING to do with marketing.
Russell Williams
> Super CCD and interpolation are marketing
(I posted this info a couple of weeks ago in another thread where
the same outraged statements were made).
As others have noted, Fuji's SuperCCD requires some kind of
processing to produce a rectangular pixel array since it doesn't start off
that way. Just as the interpolation from a Bayer pattern can produce
much better results than the naive claim of 1/3 the actual number of
pixels on the CCD, the Fuji process should be able to produce
better results than the number of pixels on their CCD. Surely not
2X, but it's not a complete fabrication either. It's also possible that
better algorithms have upped this number in the newest cameras.
If you had the raw sensor data, you could do both the standard Bayer
pattern processing and whatever "interpolation" Fuji is doing in the
computer, but the data you start with wouldn't look like an image if you
put it on the screen, and the "resize" or "interpolate" commands in image
processing programs would not help, because the processing is really not
like
bicubic interpolation.
However, the only way I would use megapixels -- claimed, effective,
or otherwise -- is as a guide for which cameras it would be useful
to compare against. Just look at the resolution results on DPreview
from the DImage 7, Sony F707, and Nikon 5000 -- all of which
reportedly use the same imaging chip. The 707 is as much better
than the D7 as the D7 is better than the best 4MP cameras. A useful
measurement shows how much detail you can actually record through
the lens into the final image.
DPReview reported resolution figures for the 3MP S1: 1300/1200/1000
horizontal / vertical / diagonal lines vs. 900/900/900 for the 3MP Nikon 990
and 1000/1150/1000 for the D30. This is consistent with "not 2X but not
a complete fabrication".
Russell Williams
not speaking for Adobe Systems
B.Server
<r...@nospam.demon.co.uk> wrote:
>In article <mbl16ucbm6s7eacoq...@4ax.com>, Mark Herring
><mherr...@sprynet.com> writes
>>
[...]
>Perhaps you would like to explain why the SuperCCD was awarded top
>position for TECHNICAL merit by the IEEE 18 months ago, a group of
>people who exhaled more knowledge about imaging technology in their
>morning jog than you are ever likely to learn in your entire life and
>who have NOTHING to do with marketing.
Since the IEEE is a very large group, encompassing technologies
ranging from software through electro-optics by way of semi-conductor
manufacturing processes, could you post a reference to the
beauty-contest to which you refer?
And, just for my curiousity, why are you so worked up about a
breathless marketing description of a chip?
Kennedy McEwen
<lwne...@ix.net.invalid> writes
>
>Since the IEEE is a very large group, encompassing technologies
>ranging from software through electro-optics by way of semi-conductor
>manufacturing processes, could you post a reference to the
>beauty-contest to which you refer?
>
>And, just for my curiousity, why are you so worked up about a
>breathless marketing description of a chip?
Because, as I have said on many occasions, it isn't a marketing
description of a chip. There are plenty of technical issues involved
and it results in REAL resolution gains (as shown by the data that
Russell Williams posted here twice). Also, because I am involved in the
design of imaging chips and sensors, it is especially galling to read
morons (who know nothing of the technology and understand less) refer to
the genuine technical advantages as "marketing hype" or "lies".
B.Server
<r...@nospam.demon.co.uk> wrote:
>In article <a1e56usqpp2d9jsvf...@4ax.com>, B. Server
><lwne...@ix.net.invalid> writes
>>
>>Since the IEEE is a very large group, encompassing technologies
>>ranging from software through electro-optics by way of semi-conductor
>>manufacturing processes, could you post a reference to the
>>beauty-contest to which you refer?
>>
>http://www.dpreview.com/news/0107/01070201fujifilmaward.asp
>
>>And, just for my curiousity, why are you so worked up about a
>>breathless marketing description of a chip?
>
>Because, as I have said on many occasions, it isn't a marketing
>description of a chip. There are plenty of technical issues involved
>and it results in REAL resolution gains (as shown by the data that
>Russell Williams posted here twice). Also, because I am involved in the
>design of imaging chips and sensors, it is especially galling to read
>morons (who know nothing of the technology and understand less) refer to
>the genuine technical advantages as "marketing hype" or "lies".
Since you continue to substitute invective and name calling for data,
it is a little hard to take your claims to professional involvement
very seriously. One can only hope that not everyone with whom you
disagree is a moron into whose mouth you must place words to justify
your dearly held opinions. (I did not use the term "lies". You did)
But, what the hell.
First. It comes as a surprise that you post as a reference, not the
original paper from Fuji, but rather a review from a commercial web
site that is based largely on Fuji's promotional materials. You have
read the original in your "professional capacity", have you not?
Second. Did you happen to note that the IEEE award was for a paper,
not a device?
Since the web is apparently your concept of professional research, I
did a couple of web searches hoping to find three things.
[ 1] Information about the Kosonocky Award, its history, other
winners, and the criteria used in submission/selection.
[2] The text of the Fuji paper that won the award.
[3] Some indication of the discussion of the paper, award, and
technology.
[ 1] The Fuji paper was the second award. The award has only existed
since 1999. A search at IEEE.org does not return any (none, nada)
hits for "Kosonocky Award" so the word is not getting around very
quickly.
[ 2] So far, no sign of the original paper.
[3] Around 130 hits for my search, of which 127 were either Fuji press
releases, advertisements, and other publicity or journalists writing
about the technology/award and largely basing their writing on the
Fuji PR stuff. . One (1) was a reference to the bio of a Dutch
researcher who was the chair of the committee who originated the
award. One (1) was was a reference in the minutes of a Chicago-area
IEEE chapter meeting. One (1) was a reference to an entirely other
matter. It appears to me that the PR angle somewhat outweighs the
technical exposition...
As a self proclaimed member of the technical elite, you are no doubt
familiar with the practice of evaluating research papers by the number
of times they are cited in other researcher's efforts. So far, at
least in your medium of choice, its pretty thin. The score looks like
PR by better than 100:1.
This technology with which you are so smitten may indeed be the most
significant thing that has happened to light since Genesis, but we are
not likely to discover that from your habits of discourse.
-and you still seem to have an excessive attachment to it-
Kennedy McEwen
<lwne...@ix.net.invalid> writes
>
>Since you continue
I suggest to look up the definition of "continue" if you insist on the
use of accurate terminology.
>to substitute invective and name calling for data,
>it is a little hard to take your claims to professional involvement
>very seriously. One can only hope that not everyone with whom you
>disagree is a moron into whose mouth you must place words to justify
>your dearly held opinions. (I did not use the term "lies". You did)
>
Indeed I did use the term "lies" - however, you chose to consider it was
directed specifically at you. Why? Guilty conscience or just continued
stupidity in the face of the facts? Look further back in the thread and
you will find no shortage of misguided individuals using that exact
term. Anyone who claims that a system which produces exactly the
results that established physics predicts it to yield is misleading,
deception or lying IS a moron, and my personal opinion of them is moot!
>But, what the hell.
>
>First. It comes as a surprise that you post as a reference, not the
>original paper from Fuji, but rather a review from a commercial web
>site that is based largely on Fuji's promotional materials. You have
>read the original in your "professional capacity", have you not?
>
I cannot be certain of reading that specific paper - I would need to
look out the actual publication in my company's library to be sure.
However I probably have, and have certainly read several Fuji papers in
recent years which have addressed this approach, being similar in at
least one aspect to a solid state implementation of a
semi-optomechanical solution I hold the patents on.
However it should be no surprise to you that none of the papers I have
read refer to any specific awards. I doubt I would find such a
statement in any edition of any refereed technical journal. Instead I
referred you to a reporting organisation for such information - an
organisation such as Digital Photography Review which, unlike your
claims, is not based largely on Fuji promo material but considers most
of the leading competing systems in the marketplace.
>Second. Did you happen to note that the IEEE award was for a paper,
>not a device?
>
Indeed. It was for the BEST paper representing "significant advancement
in solid-state image sensors" in a refereed technical publication of the
IEEE. What more acknowledgement of the technical merit for their claims
do you want - the paper describing specifically the SuperCCD won THE
award for best technical advancement!
You know the claims, you have read of the accolades from the scientific
community, you have read the measured results of independent test houses
and still you consider it to be deception. You are looking more like an
ostrich with every post you write.
>Since the web is apparently your concept of professional research, I
>did a couple of web searches hoping to find three things.
> [ 1] Information about the Kosonocky Award, its history, other
>winners, and the criteria used in submission/selection.
> [2] The text of the Fuji paper that won the award.
> [3] Some indication of the discussion of the paper, award, and
>technology.
>
Clearly the web is YOUR concept of research! You asked for a reference
to the award - I provided you with one you could easily access. If you
want to read the paper I suggest you procure the relevant copy of the
IEEE Technical Journal it was published in. I do not think you will
find the information you seek on the web as most technical publications
are not, at least freely, available there. Most technical libraries
will be able to get you a copy or you can buy an original yourself
directly from the IEEE publications. I suggest you widen your scope of
research if you want the technical information at first hand - or begin
to learn a little of the imaging physics involved so that you are in a
position to work it out for yourself! As I have already mentioned, the
resolution benefits of the SuperCCD have been long known, but Fuji have
been the first to implement them directly in the sensor itself.
>[ 1] The Fuji paper was the second award. The award has only existed
>since 1999.
Hardly surprising since it is a *biannual* award given in honour of
Walter Kosonocky who died unexpectedly only a year or two earlier! That
tends to be the way of these things - few awards are made in honour of
people who are still alive and, therefore, quite capable of winning them
themselves! Presumably you consider the value of Nobel Prizes to be
equally lacking since they also only came into existence five years
after Alfred Nobel's death as well!