SD9 sample images on IR

[Ladehaug]

It is certainly evident that the SD9 can capture more detail per
picture element. BUT, the d60 and S2 have more picture elements to
work with so it is able to capture comparable detail overall.

IMO resolution measured off the a chart has an objective meaning.
All lines must be visible. So blow up each the Sd9, D60 and S2 res
chart and tell me where the last point is that you can see 9 black
lines with 8 contained white lines. That is your absolute
resolution.

When I do this I get:
D60 1500
SD9 1525
S2 1600

That much I can agree with. Down there by by the number "20"
one gets the impression that there are five lines, not 9, so it is
clear that something isn't quite right ... because there should be
9, not 5.

It is not possible with the current resolution (picture height is 1512, therefore limiting the max resolution to approx 15 on the scale). Check the width (number of pixels) at the 20 mark. There is simply not enough pixels to render 9 separate lines within 14 pixels. It will require at least 18 pixels to render 9 separate lines.

Hardly significant difference. Anything beyond this point is noise,
what you both seem to arguing is the subjective quality of that
noise.

But it does seem odd that it is vertical noise, in the shape of black
lines, just like the ones above. Just teasing, I say that seems like
a pretty strange coincidence.

But you are right. Four of the lines disappeared somewhere along
the way. Something is not quite right.

They will disappear one by one in an "phase correlation" / aliasing process. It fully depends on how much each of the photocells are seing: If the first photocell sees 1 whole line and also 40% of the next line, then the next photocell is seing the rest of that line (60%), and also all or a part of the next line.. etc. etc.. A "phase correlation" is happening as long as the distance between one or more lines doesn't even up with the distance between photocells. This will lead to aliasing at certain frequencies when adjacent sensors sees information that are ending up being the same color. Uuuh - this is difficult to explain! Should have made some drawings etc.., but I'll let it go for now

For a subjective test compare the davebox images of the SD9 against
either of the others after ressing up the images to a new size. I
chose 4500x3000 so all images would have to be resized. Compare
detail and image quality and tell me which one you prefer and why.

Really I would like to know if we are seeing something different.

It's getting late here. I promise to look tomorrow.

 

[Ferenc]

Peter:

On the SD9, the reading is WAY below 1500. In fact, past 1200, when spatial frequency increases significatnly, the SD9's resolving power exhibits erratic behavior, where in some sections the sample is blurred, and then it comes back to almost entirely visibility, and then it quickly becomes extinguished.

In any 6 MP sample from the D100, D60 or S2, these reading as STABLE up to 1600+ (on the S2, which shows the highest). In fact, if you take Phil's RAW sample from the S2 @12.12 MP, you will clearly see spatial resolution being pushed up all the way to 1800 and beyond!!! This equates to a full 15% increase in net resolution, which, in terms of Megapixels, equates to 8-9 Mp.

The SD9 does not deliver (yet) on the hyped doubled resolution that it was supposed to do. But, eventually, it will definitely will. It is just a matter of time and technological maturity.

Kind regards,

Ferenc

It is evident that the SD9 image is able to capture and
differentiate more details per picture element than the D60. The
result is SHARPER pictures with MORE details from the SD9 than the
D60.

It is certainly evident that the SD9 can capture more detail per
picture element. BUT, the d60 and S2 have more picture elements to
work with so it is able to capture comparable detail overall.

IMO resolution measured off the a chart has an objective meaning.
All lines must be visible. So blow up each the Sd9, D60 and S2 res
chart and tell me where the last point is that you can see 9 black
lines with 8 contained white lines. That is your absolute
resolution.

When I do this I get:
D60 1500
SD9 1525
S2 1600

Hardly significant difference. Anything beyond this point is noise,
what you both seem to arguing is the subjective quality of that
noise.

For a subjective test compare the davebox images of the SD9 against
either of the others after ressing up the images to a new size. I
chose 4500x3000 so all images would have to be resized. Compare
detail and image quality and tell me which one you prefer and why.

Really I would like to know if we are seeing something different.

Peter

 

[astroman]

Say U,

You mind if ask what kind of scope and tracking drive/software your friend used on that shot? Really sharp!

G

Still, you are right about the higher ISO and the longer exposures.
I have my imaginary setup for astrophotography all put together in
my mind, and the long exposures would be a dream to be able to
achieve. This is what I have in mind, as done by Joe Jones, D60:

Really I still think my imaginary dollars would be in the D60,
those clean 4min exposures knocked my socks off.

Peter

--

Ulysses
Repository of Some of My Stuff
http://www.imagestation.com/album/pictures.html?id=4291269101

I'm an uncle!!!

 

[Peter G]

Yes, I noticed this but was trying to give the SD9 the benefit of the doubt. When you get near the limits of needing 2 pixels to resolve the line pair aligment becomes crucial. It did come into alignment right around 1500 so I called it there.

But in reality that is an advantage to the bayer camp since they are throwing more pixels at it, the bayer cameras with more pixels are little less subject to alignment issues.

For the sake of argument I am agreeable to saying that the SD9 has similar resolution to the 6MP cameras. It is not different enough either way to be a differentiator.

But when you look at the highlight issues, noise and low light performance, the current cameras do look to be ahead.

Peter

On the SD9, the reading is WAY below 1500. In fact, past 1200, when
spatial frequency increases significatnly, the SD9's resolving
power exhibits erratic behavior, where in some sections the sample
is blurred, and then it comes back to almost entirely visibility,
and then it quickly becomes extinguished.

In any 6 MP sample from the D100, D60 or S2, these reading as
STABLE up to 1600+ (on the S2, which shows the highest). In fact,
if you take Phil's RAW sample from the S2 @12.12 MP, you will
clearly see spatial resolution being pushed up all the way to 1800
and beyond!!! This equates to a full 15% increase in net
resolution, which, in terms of Megapixels, equates to 8-9 Mp.

The SD9 does not deliver (yet) on the hyped doubled resolution
that it was supposed to do. But, eventually, it will definitely
will. It is just a matter of time and technological maturity.

Kind regards,

Ferenc

It is evident that the SD9 image is able to capture and
differentiate more details per picture element than the D60. The
result is SHARPER pictures with MORE details from the SD9 than the
D60.

It is certainly evident that the SD9 can capture more detail per
picture element. BUT, the d60 and S2 have more picture elements to
work with so it is able to capture comparable detail overall.

IMO resolution measured off the a chart has an objective meaning.
All lines must be visible. So blow up each the Sd9, D60 and S2 res
chart and tell me where the last point is that you can see 9 black
lines with 8 contained white lines. That is your absolute
resolution.

When I do this I get:
D60 1500
SD9 1525
S2 1600

Hardly significant difference. Anything beyond this point is noise,
what you both seem to arguing is the subjective quality of that
noise.

For a subjective test compare the davebox images of the SD9 against
either of the others after ressing up the images to a new size. I
chose 4500x3000 so all images would have to be resized. Compare
detail and image quality and tell me which one you prefer and why.

Really I would like to know if we are seeing something different.

Peter

--

 

[Ladehaug]

But at 1200-1400 it cannot resolve all 9 lines. You are not sure
how many lines there are. 8? 9? 10?

You are right. When the lines are 1 pixel sized (width), there isn't much space to "shift" the line to the next pixel position without touching the data from an adjacent line. This is the price to pay for rendering sharp details using a limited resolution.

More accepted to report no information?

Yes, if you wish to avoid aliasing errors.

Ok, if in this certain case "no information" means grey I'm with you. Well, as I've said before, I don't see aliasing as a problem. In practically all sample pictures I've seen from the SD9, I don't see this as a problem. Therefore I would not interpolate / soften / blur the image just to avoid aliasing. I would rather keep the details, with "aliasing errors" or not. Let's say a leaf from a tree have 9 lines and it spans 20 pixels on the image. I would like to see 5 aliased lines on the leaf rather than no lines at all.

What erroneous data? The
sensor cells sees what it sees!

But you cannot be sure exactly what it is seeing.

If a specific sensor cell sees 60%
of the line (black) and 40% of the "background" (white), then what
should it report? The adjacent pixel perhaps see 60% background and
40% of another black line.

It may report both as white, both as black, one one white and one
black. But which of the 4 possibilities is correct? You can never
tell. You will also get different answers different times based on
a lot of other factors.

The sensor does not act as a perfect point sampler!

If the photocell "sees" 60% of the black line and 40% of the white background, then the resulting color it actually sees, would be a mix of these colors. In this case, this would ideally be a grey color (CA will have some influence, so will the focus point and the quality of the lens. No lense is good enough to make the sensor a perfect point sampler.

Or should that specific sensor cell go asking the
neighbouring cell, asking "Should I adjust my color, so it becomes
more similar to yours?", and then report some kind of mix?

Well, that would be digital filtering. More common is to use an
analog filter e.g. anti-aliasing filter. That's not primarily a
photographic term, you know.

One can't avoid aliasing without blurring out details.

However, those "details" have very low confidence. Is it better to
present them as accurately measured data when they are not? Don't
confuse precision and accuracy.

Details have low confidence? Yes, we can count 5 lines at the 20 mark. Certainly, in reality there are 9, so the picture is lying to us. On the D60 image, there are no lines to be seen at the 20 mark, so when we look at it, we just have to assume that there are no further details within this area. Besides it is even colored light blue on the D60 image. How can we be confident at all?

If aliasing
will be a problem, then simply soften/blur the image in Photoshop.
I can't see this as a problem in practically all sample pictures
I've seen so far.

Well, we have all this sampling theory, but the real issue is
perceptual. Which makes a better photograph? Which happens more
often and is more noticable: aliasing errors or loss of "detail"
when anti-aliasing? This may end up as personal preference. Or
there may be other contributing/mitigating factors (e.g. print vs.
monitor viewing.) Once you know what aliasing errors look like, you
can find them in many SD-9 images. (Example: the various metal
rings in IMG00137.)

I'm with you on this. Photographic preference is an important issue. Personally, as an example, I would like to take more detailed nature shots. In this case, more details are more important to me than thinking of aliasing errors in patterns etc.

Geir Rune

 

[David SL]

Any resize puts the resized image at a bit of a disadvantage
compared to an unresized native image. Some softness is introduced.

Exactly, but when the smaller dimensions of the Sd9 image 2268x1512 are sized to 4500x3000 the magnification (and likewise degredation) it is experiencing is a bit greater than the jump of the D60 pixel dimensions of 3072x2048 to the same size.

I think Ulysses point is(U, correct me if I am wrong on this.), that the fact that you are comparing a 3mp image at such magnification to a 6mp image to see a hint of difference speaks volumes on just how good that 3mp image really is.

For a more fair comparison try the same blowup routine with a D30 image.

Regards,

Plus by going bigger I can see the differences more readily.

Peter

Nope, I upsized them both up to the same larger
resolution(4500x3000 for the full image). I think it is the fairest
way to compare images of differing sizes.

--

Ulysses
Repository of Some of My Stuff
http://www.imagestation.com/album/pictures.html?id=4291269101

I'm an uncle!!!

--

 

[Erik Magnuson]

I would rather keep the
details, with "aliasing errors" or not.

Can't argue with personal preference.

The sensor does not act as a perfect point sampler!

If you kick the contrast up enough it does! ;-)

--
Erik

 

[Group Idiot]

For a more fair comparison try the same blowup routine with a D30
image.

The SD-9 isn't competing with the D30, a long-discontinued camera. It's competing with the D60/100/S2.

Brendan
--

Newbie requests inspection and critique of new photos at: http://www.pbase.com/bgetchel/root

 

[Ferenc]

The closest image you can compare is the Resolution Chart (due to the particular way it needs to be captured).

Now... why don't you do the evaluation on your side, just like you suggest, and come back and post the results, so we can all see?

Enjoy,

Ferenc

But I would res both up to say 20 MP. It is hardly fair to compare
one image at its native resolution and another that will suffer
processing softness to get there.

Peter

Then you will be able to see the actual differences...

http://www.dpreview.com/reviews/fujis2pro/page12.asp

Check out the Raw vrs Jpeg part near the bottom. The S2 already
does better on res charts than the other 6MP cameras in JPG mode.
The performance in Raw is even more impressive.

Then again, it really didn't really show that large difference in
images, but it certainly does the best Res chart.

Really I still think my imaginary dollars would be in the D60,
those clean 4min exposures knocked my socks off.

Peter

--

 

[Robert P Miller]

Yeah, me too. I'm not committed to anything (did I tell you I sold
my "other" camera?). So I'm free as a bird and checking a little
bit of everything now.

Ulysses, you're Sonyless at present? Say it ain't so Joe, just say it ain't so!

I'm crusin' the forums too and have checked out a few of the DSLRs at the Javit's photo show. Did you get a chance to go?
Best,
Robert

 

[Ulysses]

Nope, I don't mind. But credit goes to Joe Jones.

Have a look at just some of his gear. Joe is serious about his scopes and owns several. You can consider the Astro-Physics refractor a "serious" telescope. This is Joe's 6" refractor.



His D60/astronomy pbase galleries are located here:
http://www.pbase.com/jayseejay/d60_digiscoping&page=2

Wanna know more about Astro-Physics scopes, go here:
http://www.astro-physics.com/

Say U,
You mind if ask what kind of scope and tracking drive/software your
friend used on that shot? Really sharp!

--

Ulysses
Repository of Some of My Stuff
http://www.imagestation.com/album/pictures.html?id=4291269101

I'm an uncle!!!

 

[Ulysses]

Ulysses, you're Sonyless at present? Say it ain't so Joe, just say
it ain't so!

It's true, guy. I'm a free man!! And I'm checkin' out the field.

I'm crusin' the forums too and have checked out a few of the DSLRs
at the Javit's photo show. Did you get a chance to go?

I wanted to, but I didn't have the time to get over there. I hear, though, that it wasn't all that great. So....

--

Ulysses
Repository of Some of My Stuff
http://www.imagestation.com/album/pictures.html?id=4291269101

I'm an uncle!!!

 

[Veniamin Kostitsin II]

i really do.

I noticed on imaging resource that they've published their sample
box image from a production sd9 they just received in, I must say
the level of detail is impressive in such a "small" size. Hope Phil
doesn't mind me posting the info to 'another' website (though Phil
is so cute when he gets mad

Rich

--
Phil Askey
Editor / Owner, dpreview.com

--
Phil Askey
Editor / Owner, dpreview.com

--
Veniamin Kostitsin II
http://www.digitalimage.at/

 

[Robert P Miller]

I wanted to, but I didn't have the time to get over there. I hear,
though, that it wasn't all that great. So....

Well, it depends upon what great is...... It was a gear show to be sure. My dad and I got together and made a trip of it with dinner and such. We also met some folks selling tutorials on Photoshop. Their cd ROMs impressed us enough to pick one up. Dad got to complain to HP about his repeated scanner trouble and their repair QC. And then there was the hands on with the cams.......

I gotta tell ya, half (2) of the cameras I went to see were still the prototype models and as such cannot be judged by me in any but the most cursory manner. These were the S9 and 14n.

The viewfinder in the S9 was nice in that it was large and gave a wider frame from which to see. The outer area ,was masked off darker so as to bring forth the center "sport finder" application which they had chosen to favor. This was in Contrast to the S2 which had an adequate but small viewfinder. The real winner on this front was the 1Ds and 14n (no surprise). They offered large bright Viewfinders with true wide angle ability showing off its prowess at last!

This is my main reason (save for resolution, as I love landscapes) for ordering the 14n. I have keen interest in the Foveon/ sigma concept, but NEED true wide angle now. Also, the Kodak has upgradeable software and jpgs which can be controlled like raw for way more post control.

I am going to miss the live LCD and flip body of the 707, but am keeping that camera too. It is too cool of an image machine to let go of yet........
Gotta sell some stuff to make all this work though!

Olympus' booth was a bit empty compared to last year, at least from the times we were there each year (third day). I spoke to a rep about the 4/3 camera as I did not see a prototype. He was quiet and reflective when he said they would be successful, but not necessarily in the market place. I think he felt as many of us do, it is a sound concept, but too late perhaps in the current market place. I mentioned Oly's partners in this endeavor, and he seemed hesitant to commit to any comment. Perhaps all is not well in Olydak land..........
And then the 1Ds, uh. we should all own one just for the AF.

It aint gonna happen because of the forced marketing tier that exists. It is truly a shame that decent Af is only on the "Pro" models. It exists that way in the film world, so the powers that be stick to that model. I would be ok with that IF the top end models , which cost a gazillion, would have a shelf life longer than a canned ham. But I digress...
Hey, the Sony AF was not too bad was it? Love those lasers!

Gee, this ended up being a report on the show. Hmm, perhaps I should post it in the News topic area?

Any ways, the show was what you make of it. For my dad, he's doing photography now with a vengence. At 73 he now shoots with my old E-20, prints with his new Compaq computer and sells some of the prints to those he photographs. He loves the total control. Many of us never had that . Viva digital!

Happy Uncledom! I am a recent father and man, my little girl is the most photographed person in my life. Chasing kids is why we need great AF!
Best,
Robert

 

[Peter G]

Do you wish to suggest a better way to compare them? Comparing to a discontinued camera is rather pointless. Everyone already knows it should crush a D30. It makes more sense to compare with the competition.

I think my comparison is quite valid, it introduced no artifacts not present in the original images and allows us to easily see the detail.

While there is only a hint a difference in the detail present, there is a world of difference in the problems with the colour artifacts in the Sigma image, that has nothing to do with detail. It suffers colour bleed and violet blooming.

Great technology for the future, I think the 1.0 implementation is a pass.

Peter

Any resize puts the resized image at a bit of a disadvantage
compared to an unresized native image. Some softness is introduced.

Exactly, but when the smaller dimensions of the Sd9 image 2268x1512
are sized to 4500x3000 the magnification (and likewise degredation)
it is experiencing is a bit greater than the jump of the D60 pixel
dimensions of 3072x2048 to the same size.

I think Ulysses point is(U, correct me if I am wrong on this.),
that the fact that you are comparing a 3mp image at such
magnification to a 6mp image to see a hint of difference speaks
volumes on just how good that 3mp image really is.

For a more fair comparison try the same blowup routine with a D30
image.

Regards,

Plus by going bigger I can see the differences more readily.

Peter

Nope, I upsized them both up to the same larger
resolution(4500x3000 for the full image). I think it is the fairest
way to compare images of differing sizes.

--

Ulysses
Repository of Some of My Stuff
http://www.imagestation.com/album/pictures.html?id=4291269101

I'm an uncle!!!

--

 

[Peter G]

I have done this with the 12mp res chart from the RAW file in Phils review, it is a SIGNIFICANT improvement over the other 6mp camera, easily and cleanly going to 1800-1900 lines.

I also compared the the bridge photos in S2 and D100 review (to throw the D60 into the mix).

I used the 12mp raw generated image from the s2. Conclusions:

I see a clear resolution advantage of the S2 over the D100, but I also see a similar advantage for the D60 over the D100. If I had to choose between the D60 and S2 I would give the edge to the S2, BUT it is nowhere near as significant as the res chart would indicate. So I agree that it does capture more resolution, just not to the extent that you claim. Not 6MP vrs 8MP you claim, not for real images anyway.

Certain algorithms can detect the monocrhome/high contrast nature of the res chart and take advantage of that (if you know you are shooting B&W subject you can make better correlations).

I just don't see that dramatic an advantage in images, though there still is one.

I will argue no further on this, since we on seem to differ on the degree of the advantage.

Peter

Certain algorithms can extract more detail from things like monochrome

Now... why don't you do the evaluation on your side, just like
you suggest, and come back and post the results, so we can all see?

Enjoy,

Ferenc

But I would res both up to say 20 MP. It is hardly fair to compare
one image at its native resolution and another that will suffer
processing softness to get there.

Peter

Then you will be able to see the actual differences...

http://www.dpreview.com/reviews/fujis2pro/page12.asp

Check out the Raw vrs Jpeg part near the bottom. The S2 already
does better on res charts than the other 6MP cameras in JPG mode.
The performance in Raw is even more impressive.

Then again, it really didn't really show that large difference in
images, but it certainly does the best Res chart.

Really I still think my imaginary dollars would be in the D60,
those clean 4min exposures knocked my socks off.

Peter

--

--

 

[Peter G]

Here is a clip from the res chart in the DaveBox image:

SD9:



D60:



Arguing about which one has more detail in the alaising seems like looking for Potatoes that have faces in them.

If you look at the Diagonal 3 section, you will note that the D60 renders it as diagonol, the SD9 registers it as broken vertical lines.

Section 5 is filtered out on the D60 to grey. The SD9 renders it as broad diagonal stripes tilting in the wrong direction.

IMO these are similar, in fact the d60 does a slightly better job since it renders the 3 section correctly. After that neither convey any real information, certainly nothing I would call detail.

Peter

I would rather keep the
details, with "aliasing errors" or not.

Can't argue with personal preference.

The sensor does not act as a perfect point sampler!

If you kick the contrast up enough it does! ;-)

--
Erik

 

[A Anderson]

Just a quick look at the resolution chart and I see something like 1600 max horizontal and 1200 maybe 1300 vertical ?

What does everyone else see?

 

[Peter G]

I guess we would all need to agree on a definition for what resolution we are reading.

The last place I can see all 9 lines is just a bit over the 1500 mark for both the horizontal and the vertical on the SD9. But there are spots before that where all 9 are not visible.

BTW if you are into photographing RES charts I suggest the Fuji S2, as it clearly wins that contest:

http://img.dpreview.com/reviews/FujiS2Pro/Samples/RAW/DSCF0057_1.jpg

Peter

Just a quick look at the resolution chart and I see something like
1600 max horizontal and 1200 maybe 1300 vertical ?

What does everyone else see?

 

[A Anderson]

aaah, yes I am fully aware of the S2. Very impressive indeed.

The last place I can see all 9 lines is just a bit over the 1500
mark for both the horizontal and the vertical on the SD9. But there
are spots before that where all 9 are not visible.

BTW if you are into photographing RES charts I suggest the Fuji S2,
as it clearly wins that contest:

http://img.dpreview.com/reviews/FujiS2Pro/Samples/RAW/DSCF0057_1.jpg

Peter

Just a quick look at the resolution chart and I see something like
1600 max horizontal and 1200 maybe 1300 vertical ?

What does everyone else see?

--
Sony DSC-F707 Amateur Photographer