SD14 JPEG interpolation

[LMC54]

I wonder what interpolation method will be used.

I would suspect as far as results go they will want something
conservative in effect like Bicubic or Lanczos. Rather than some
edge enhance s-spline techniques.

Other than that they will want something that isn't to
computationally intensive.

Why does it matter? Do you really think they are going to have some
super expansion method so it will be worthwhile to do it on camera,
since you won't easily duplicated off camera??

This is nothing more than a marketing move to have a mode that
lines up output pixel count with advertised pixel count.

As intensive as Bayer interpolatioin goes I don't see why not a cool method for Sigma.

--
http://www.fredmiranda.com/hosting/showgallery.php?ppuser=235&cat=500
http://www.pbase.com/lmc54/sd10

 

[mauricm]

from the sensor and a JPEG is rendered from that how is that upsizing?
What difference does it make if the data is collected as in Bayer
from horizontally arranged photosites on a larger sensor or from a
smaller sensor with stacked photosites?

It is an essential difference. To be able to (potentially) extract
more spatial information then the color sensors must be in
different positions. If they are in the same position, then
upscaling is simply .... upscaling .. no information is added ...
it just takes more space.

So if you ...

(1) ... put the three color detectors in the same position, you get
more accurate color information.

(2) ... put them at different positions, you can extract more
spatial luminace resolution.

--
Roland
http://klotjohan.mine.nu/~roland/

You said it Roland. That was absolutely correct. Converting sensor cells to pixels is not so easy like someones like to advice us.

 

[Kendall Helmstetter Gelner]

It is sized to give it 14MP as it is being advertised as a 14MP
camera. But this is a waste of space mode with no real purpose
other than marketing.

I agree for most uses it will probably make more sense to select the high resolution JOPG.

However if you were in the situation where you were going to use a kiosk for printing and you didn't know how it did interpolation, it might be better to let the camera do a good job instead of risking an enlargment that only did a stright rescaling with no resampling. Kiosks and computerless printing solutions are becoming more common, and I would rather drop off a 14MP JPG with a photofinisher than an un-resized image.

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmasd9/user_home

 

[Roland Karlsson]

I always think of Bayer as performing at about half the rated
count. When I say we will never get a satisfactory solution, I mean
one that everyone can agree on.

The main problem when SD9 and SD10 was introduced was that at all sites where you searched for cameras (to buy or look for review) was that the SDx cameras was sorted among 3MP and up cameras while the competition was sorted among 5MP and up.

Then if the SD10 would have been marketed as a 7 MP Equiv camera, then it would have been in the right category. Everyone would have been satisfied. The sun would shine and the grass greener.

Now Foveon/Sigma choose the hard way instead - trying to convince the world it WAS a 10 Mpixel camera. But the world was stubborn and refused - and it will do it again. Then it is rainy, thunder and polluted lakes.

--
Roland
http://klotjohan.mine.nu/~roland/

 

[Roland Karlsson]

However if you were in the situation where you were going to use a
kiosk for printing and you didn't know how it did interpolation, it
might be better to let the camera do a good job instead of risking
an enlargment that only did a stright rescaling with no resampling.
Kiosks and computerless printing solutions are becoming more
common, and I would rather drop off a 14MP JPG with a photofinisher
than an un-resized image.

OK .... you managed to come up with a potential usage.

In practice I dont think it holds though - for two reasons

(1) you usually dont make high quality giant printings that way.

(2) if you do - the kiosk probably upscales just as fine as your camera.

But otherwise - 3 points for a nice try

--
Roland
http://klotjohan.mine.nu/~roland/

 

[Kendall Helmstetter Gelner]

As the X3 sensor outputs 4.66 M color tripples, where all three
samples are in the same spatial location - then it can only be
ordinary upscaling. No extra spatial information exists to increase
the resolution.

That Sigma/Foveon have chosen to add the 14 Mpixel JPEG mode is
totally unneccessary for the photographer. Its just there to
confuse the discussions whether it is a 14 Mpixel camera or not
IMHO.

Of course it's not really adding any more information. It's just expanding on the greater amount of information that is already there.

But are there not cases today where Sigma users upscale images before sending them off somewhere? After all, SPP does have a double-size mode... sometimes for printing, or possibly for other reasons. It also provides a file that you can more directly compare against a Bayer image as you either have to shrink the bayer image to account for a lack of sharpness or increase the X3F image to get a similar view at the 100% level.

It is those same cases where the double size JPG can be of use because you can use the JPG from the card directly.

And of course I personally enjoy how it sticks a wheel in the spokes of those who repeatedly and incorrectly claim the resolution is properly described as 4.7MP. Of course 14MP using the classical Bayer definition is not right either, but it puts the shoe on the other foot forcing people to explain why it's not 14MP instead of forcing Sigma users to explain why it's not 4.7MP. And it's probably closer to correct to say it's a 14MP camera rather than a 5MP camera, if you were going strickly by the numbers.

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmasd9/user_home

 

[Erik Magnuson]

However if you were in the situation where you were going to use a
kiosk for printing a

Now all Sigma needs is a "Direct Print" button... ;-)

--
Erik

 

[Kendall Helmstetter Gelner]

My guess is that a crop 1.3, 8x3 Mpixelx Foveon camera blows the
competition out to sea. It might not cost $1600 but rather $2500 -
but it would not matter. It would be a success.

Probably more like $3000, at least, due to the larger chip size. Increasing chip size is not a linear increase in cost.

Furthermore it would have been a pretty annoying step for a lot of Sigma users who bought DC lenses (which take around a max of a 1.5 crop).

In time perhaps we'll see a FF or 1.3x Sigma, but it would have been too big of a step for now and alienated too many from a very loyal user base. Give them some time work up the ladder! Success can come with the current camera, because we get improved resolution along with a lot of user friendlier features.

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmasd9/user_home

 

[Roland Karlsson]

Of course it's not really adding any more information. It's just
expanding on the greater amount of information that is already
there.

Some think it does ... and thats the main reason IMHO for this mode ... to confuse the issue.

And of course I personally enjoy how it sticks a wheel in the
spokes of those who repeatedly and incorrectly claim the resolution
is properly described as 4.7MP.

If you remove "incorrectly" I think some (including me) would take you more seriously. SD14 has a maximum resolution limit capped by 4.7M three color samples. Those sample are called pixels. And thats what determines the maximum possible spatial resolution.

Of course 14MP using the classical
Bayer definition is not right either, but it puts the shoe on the
other foot forcing people to explain why it's not 14MP instead of
forcing Sigma users to explain why it's not 4.7MP. And it's
probably closer to correct to say it's a 14MP camera rather than a
5MP camera, if you were going strickly by the numbers.

Have you seen my proposal that you could call a SD14 a 10 MPixel Equiv camera? How about that?

--
Roland
http://klotjohan.mine.nu/~roland/

 

[Kendall Helmstetter Gelner]

(1) you usually dont make high quality giant printings that way.

Well, I was thinking more for 8x10's perhaps.... or signficant crops.

(2) if you do - the kiosk probably upscales just as fine as your
camera.

Actually that would make a really interesting test. I should try going to a kiosk someday and compare one resized picture printed alongsie an un-resized picture where the unresized on would be scaled to the same size... I never have placed a lot of trust in the editing ability of those kiosks, or with photo services in general.

But otherwise - 3 points for a nice try

Well I thought it was pretty good with such seemingly barren material as why you'd upscale a JPG in camera... thanks.

Actually isn't it at least a more useful feature than HP's slimming mode? There are sillier things that can be done to an image in camera!

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmasd9/user_home

 

[Roland Karlsson]

Probably more like $3000, at least, due to the larger chip size.
Increasing chip size is not a linear increase in cost.

Furthermore it would have been a pretty annoying step for a lot of
Sigma users who bought DC lenses (which take around a max of a 1.5
crop).

In time perhaps we'll see a FF or 1.3x Sigma, but it would have
been too big of a step for now and alienated too many from a very
loyal user base. Give them some time work up the ladder! Success
can come with the current camera, because we get improved
resolution along with a lot of user friendlier features.

Fully agree. Its just that the current strategy seems to be a mixture between hi-end and low-end that makes the camera into a crocofant (crocodile/elefant). Neither this nor that. It is priced as a hi-end 10 Mpixel camera, but is no hi-end camera.

--
Roland
http://klotjohan.mine.nu/~roland/

 

[Roland Karlsson]

Actually isn't it at least a more useful feature than HP's slimming
mode? There are sillier things that can be done to an image in
camera!

uh eh grrrr ... why do they do such stuff?

Putting lots of stupid things in the cameras. If I could take my *istDS apart and localise all the stupid modes and other stuff and then remove those and put the camera together again - adding ISO in the view finder - a spot metering button - then I would do it. And feel better.

--
Roland
http://klotjohan.mine.nu/~roland/

 

[Kendall Helmstetter Gelner]

If you remove "incorrectly" I think some (including me) would take
you more seriously. SD14 has a maximum resolution limit capped by
4.7M three color samples. Those sample are called pixels. And thats
what determines the maximum possible spatial resolution.

I am kind of in agreement along those lines, the problem is that you also need to "correctly" state a Bayer sesnor currently rated at 10MP is really a 5MP camera - sort of, because the AA filter confuses the issue even more along with the extra pixels of different colors. Until then it is more incorrect to say the SD-14 is 4.7MP because you are then giving out a number that will be incorrectly used for comparison purposes.

Have you seen my proposal that you could call a SD14 a 10 MPixel
Equiv camera? How about that?

Yes (I've seen the idea) but you have to declare a range really, you can't have a single number as I've noted in other posts. There is some minute detail a Sigma sensor will capture or represent that a bayer sesnor will not around the limits of single pixels, or rather may or may not get depending on the colors involved and the strength of the AA filter used. It's in that grey area at the edge of "resolvement" (a fine word that I have just created because nothing real seemed to fit as well) where the defintiions get tricky.

It is a much better number to use as a starting point, I just feel it's still a little too inaccurate to use outright (or in marketing materials).

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmasd9/user_home

 

[Kendall Helmstetter Gelner]

Fully agree. Its just that the current strategy seems to be a
mixture between hi-end and low-end that makes the camera into a
crocofant (crocodile/elefant). Neither this nor that. It is priced
as a hi-end 10 Mpixel camera, but is no hi-end camera.

I'm not sure about that, if the rest of the somewhat unknown aspects of camera have improved as much as the shutter reportedly has then it could be a Elephino in Kudu clothing. Whatever that means.

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmasd9/user_home

 

[Roland Karlsson]

I am kind of in agreement along those lines, the problem is that
you also need to "correctly" state a Bayer sesnor currently rated
at 10MP is really a 5MP camera - sort of, because the AA filter
confuses the issue even more along with the extra pixels of
different colors. Until then it is more incorrect to say the SD-14
is 4.7MP because you are then giving out a number that will be
incorrectly used for comparison purposes.

Yep --- but unfortunately - 14 is not better than 4.7 in this respect.

Yes (I've seen the idea) but you have to declare a range really,
you can't have a single number as I've noted in other posts. There
is some minute detail a Sigma sensor will capture or represent that
a bayer sesnor will not around the limits of single pixels, or
rather may or may not get depending on the colors involved and the
strength of the AA filter used. It's in that grey area at the edge
of "resolvement" (a fine word that I have just created because
nothing real seemed to fit as well) where the defintiions get
tricky.

It is a much better number to use as a starting point, I just feel
it's still a little too inaccurate to use outright (or in marketing
materials).

It has to be one number ... and it has to be inaccurate. To make it more complex or exact is not doable.

And the best would be that a 10 Mpixel Bayer was called a 5 Mpixel Equiv and a 4.7 full collor called 4.7. But in the next best world - both could be called 10 MP equiv.

--
Roland
http://klotjohan.mine.nu/~roland/

 

[Kendall Helmstetter Gelner]

Now Foveon/Sigma choose the hard way instead - trying to convince
the world it WAS a 10 Mpixel camera. But the world was stubborn and
refused - and it will do it again. Then it is rainy, thunder and
polluted lakes.

I agree with this thinking, which is why though Sigma needs a double-size JPG poutput to refocus the conversation and let people be a little dissapinted it's not quite 14MP instead of wondering why you'd buy a 4.7MP camera.

It's partly then a defense mechanism from the cruel world which does like to rain on parades and picnics.

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmasd9/user_home

 

[Kendall Helmstetter Gelner]

Yep --- but unfortunately - 14 is not better than 4.7 in this respect.

Ah, but why not? If you are willing to use 10 as a basis is not 14 closer to 10 than 4.7? It's seemingly less wrong, just in the other direction.

It has to be one number ... and it has to be inaccurate. To make it
more complex or exact is not doable.

And the best would be that a 10 Mpixel Bayer was called a 5 Mpixel
Equiv and a 4.7 full collor called 4.7. But in the next best world

• both could be called 10 MP equiv.

That's where I have to disagree, because I don't agree that a number has to be inaccurate or that a range is very hard to understand. People understand ranges pretty well.

Like I've said in other posts another problem is you also cannot say a 10MP bayer camera is going to capture the same detail as another 10MP bayer camera. This to me is a really serious problem because people are treating these numbers quite seriously and it's causing a lot of confusion.

That's why I can't get behind a single number, because everything is in motion with any one number you pick.

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmasd9/user_home

 

[NarrBL]

and for you, Kendall.

After all this endless noise, I decided to try a 1.4 Bicubic upsize on a SD10 image of my own.

The result: something surprising.

-- It now has many qualities a great deal like the 'best' images from a Canon 1ds, in its 'look' at full size. It is a bit sharper.

-- many details that are present but subtle at 1x become much more visible at 1.4x.

This is very interesting, and I don't think it is just my eyeglasses. Birds, antennas, and the intricate white and beige inlaid tiling of an image on the side of our Rathus (city hall) all are much more present in the picture.

-- I picked this image for some sharp, angled edges taken with the sharp 50mm prime in harsh winter afternoon sunlight, and the advantage Jay mentioned does show. There isn't much failure of aliasing, but it is smoothed to invisibility in the 1.4x version.

So - this 14 mp output choice may be as much a choice of the engineering team as any marketing. And what marketing it may do: what an interesting thing if others think it looks like Canon flagship output...

By the way, Kendall, I appreciate that you talk to Roland. It is very clearly what he wants.

Kind regards,
Clive

As the X3 sensor outputs 4.66 M color tripples, where all three
samples are in the same spatial location - then it can only be
ordinary upscaling. No extra spatial information exists to increase
the resolution.

That Sigma/Foveon have chosen to add the 14 Mpixel JPEG mode is
totally unneccessary for the photographer. Its just there to
confuse the discussions whether it is a 14 Mpixel camera or not
IMHO.

Of course it's not really adding any more information. It's just
expanding on the greater amount of information that is already
there.

But are there not cases today where Sigma users upscale images
before sending them off somewhere? After all, SPP does have a
double-size mode... sometimes for printing, or possibly for other
reasons. It also provides a file that you can more directly
compare against a Bayer image as you either have to shrink the
bayer image to account for a lack of sharpness or increase the X3F
image to get a similar view at the 100% level.

It is those same cases where the double size JPG can be of use
because you can use the JPG from the card directly.

And of course I personally enjoy how it sticks a wheel in the
spokes of those who repeatedly and incorrectly claim the resolution
is properly described as 4.7MP. Of course 14MP using the classical
Bayer definition is not right either, but it puts the shoe on the
other foot forcing people to explain why it's not 14MP instead of
forcing Sigma users to explain why it's not 4.7MP. And it's
probably closer to correct to say it's a 14MP camera rather than a
5MP camera, if you were going strickly by the numbers.

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmasd9/user_home

 

[Roland Karlsson]

Yep --- but unfortunately - 14 is not better than 4.7 in this respect.

Ah, but why not? If you are willing to use 10 as a basis is not 14
closer to 10 than 4.7? It's seemingly less wrong, just in the
other direction.

Because 14M hints at it being the number of detectors that is counted ... and we are back on square zero again. 4.7M (classic) pixels or 10M Equiv pixels are the only ones I think is OK.

That's where I have to disagree, because I don't agree that a
number has to be inaccurate or that a range is very hard to
understand. People understand ranges pretty well.

So - whats your propsed range for the SD14?

Like I've said in other posts another problem is you also cannot
say a 10MP bayer camera is going to capture the same detail as
another 10MP bayer camera. This to me is a really serious problem
because people are treating these numbers quite seriously and it's
causing a lot of confusion.

That's why I can't get behind a single number, because everything
is in motion with any one number you pick.

I fully agree. But a number is unfortunately neccessary.

And unfortunately ... now it is half past midnight in Sweden and it is bed time!

--
Roland
http://klotjohan.mine.nu/~roland/

 

[Roland Karlsson]

So - this 14 mp output choice may be as much a choice of the
engineering team as any marketing.

Nah .. a true engineer understands you can do the upscaling later.

By the way, Kendall, I appreciate that you talk to Roland. It is
very clearly what he wants.

Now I got curious. What do I want?

--
Roland
http://klotjohan.mine.nu/~roland/