Reading mode: Light Dark

SD14 JPEG interpolation

Go to last post
Oh, there's more than that, I'm afraid, and imagine that you know it.

But I just couldn't post a list, after making one. It always feels a bit of cruelty to those who love their own cameras and forms of results.

I think we have to have patience, just as with the hungers for show images.

The subtler ways in the world bring more happiness, perhaps someone said?

Kind regards,
Clive
Kendall Helmstetter Gelner said:
Also, I don't know why I passed over it before but take a look at
all of the antennas on the roof! The tops of them all just fade in
and out of existance, my own personal pet peeve for the most
annoying Bayer behaviour right after color mosiacing. It's not
even that unnoticable at 100%.

That's exactly where an image could use a little "false" detail to
let you know exactly how far the antenna extends.

The image gives the illusion of per-pixel sharpness without all of
the benefits.

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmasd9/user_home
Click to expand...
 
Kendall Helmstetter Gelner said:
I invite people to examine all crops carefully at 400% before they
submit them for revue,
Click to expand...
400% of a 5D image on a typical 100dpi monitor is looking at a 174" wide print using the most primitive upscaling technique possible. You can pixel peep any image to death on a monitor.

There are two real questions:

1. Would an SDxx image taken with a comparable lens and processing look better/worse?

2. Could you see these differences once you spray the ink on the paper.

We've done question 1 over and over for SD10 vs. 6, 8, 10 MP CFA cameras. Remember these threads?

Vs. D200:
http://forums.dpreview.com/forums/read.asp?forum=1027&message=19104654

Vs. 20D:
http://forums.dpreview.com/forums/read.asp?forum=1027&message=12347306

There is no one definitive answer. There are subtle differences, but in the end it comes down to personal preference.

--
Erik
 
Ante Vukorepa said:
Because THIS is what a Canon 5D crop looks like:



Are you telling me you seriously expect the interpolated 14MP from
SD14 will look like that at 100%?!
Click to expand...
Look at all the anti-aliasing in that picture.

The interpolated image from an SD-14 may well look rather like that, perhaps a touch less detail, but the detail that does exist will probably have more natural extinction characteritsics at the edge of "resolvement" (I word I made up but I think is pretty clear in meaning), and also without the color artifacts.

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmasd9/user_home
 
Erik Magnuson said:
Kendall Helmstetter Gelner said:
I invite people to examine all crops carefully at 400% before they
submit them for revue,
Click to expand...
400% of a 5D image on a typical 100dpi monitor is looking at a 174"
wide print using the most primitive upscaling technique possible.
You can pixel peep any image to death on a monitor.
Click to expand...
Yes but really I only zoom to 400% because my eyes are not sharp enough to really see pixel detail at 100% on a 19" monitor at 1600x1200. I am only zooming so that I can actually see the pixels that are there. Some things then jump out that I can also see if I back out again.
Member said:
There are two real questions:

1. Would an SDxx image taken with a comparable lens and processing
look better/worse?
Click to expand...
Well it would lack many (perhaps all) of the artifacts. I know it would lack the maze. I know the antennas would look better. I know the color artifacts would be gone.

I suspect we would see greater evidence of window blinds and the face would be a little more distinct.
Member said:
2. Could you see these differences once you spray the ink on the
paper.
Click to expand...
I believe the answer would be yes for at least a few of the artifacts, especially that window with the maze. The antennas are also pretty easy to judge even at 100% from a distance. The color stuff I'm less certain since it is smaller, but some of that seems to be noticable enough (in highlights especially) that it would show and be a little distracting. I would clean it up anyway if it were my image I was printing.
Member said:
We've done question 1 over and over for SD10 vs. 6, 8, 10 MP CFA
cameras. Remember these threads?
Click to expand...
Oh Yeah.
Member said:
Vs. D200:
Vs. 20D:
Click to expand...
Of course, but that was the SD-10 compared with those cameras. Now we have a camera that is more up to the task of direct comparison becaue the resolution is more in line with those cameras.

Furthermore what we have with the foveon is the case of a direct linear increase in the number of pixels availiable. In the case of the bayer cameras we cannot know for sure there is a simple linear increase in the amount of detail in an image, partly because we have no idea how strong an AA filter these cameras really use. In short the variablity of the bayer cameras in capturing data leads us to know that these cameras have more detail, but not really how much more to a precise degree.

I don't think for example that I could take an image with a D200 and a D70 where a cropped portion of the D200 image with the pixel count of the D70 had exactly the same FOV with the same lens that I would see exactly the same detail. Heck it might even be more, but I just can't say for sure.

I feel a lot more sure saying that if I took an SD-10 and and SD-14 and tried the same expiriment I would see pretty much identical results, partly because I had done something like that in the past with the x530 and seen that the results were basically the same.

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmasd9/user_home
 
Kendall Helmstetter Gelner said:
Also, I don't know why I passed over it before but take a look at
all of the antennas on the roof! The tops of them all just fade in
and out of existance, my own personal pet peeve for the most
annoying Bayer behaviour right after color mosiacing. It's not
even that unnoticable at 100%.
Click to expand...
Except that if your were to expand a Foveon image to the same degree it would suffer badly in the comparison. Here at least is a comparison.

The 10MP vs 10MP comparison. Foveon Left, bayer Right:

 
But you're missing the whole point here.

It's not a question of what happens beyond Nyquist because Nyquist frequency of a 4.6 MP Foveon just CAN'T be the same as the Nyquist of a 13MP Bayer! So all the talk about the artifacts at or around extinction has absolutely no effect on what i'm talking about and that's RAW SPATIAL RESOLUTION!

Again, i'm not trying to say Bayer has no artifacts, that it has no aliasing, that it has a better or equal per-pixel color accuracy. I'm just saying that a modern interpolation from a 13MP Bayer reveals much more spatial information that people here are giving it credit for. Which means it's plain silly to expect a 14MP interpolation from an SD14 to be of equal spatial resolution as a 13MP Bayer. 8MP, probably. 10 with quite a stretch. But for god's sake, stop expecting miracles from a "plain ole" marketing trick.

--
--------------------------------------------
Ante Vukorepa
 
Peter G said:
Except that if your were to expand a Foveon image to the same
degree it would suffer badly in the comparison. Here at least is a
comparison.
Click to expand...
That is unlike the compariosn under discussion. I am not saying "Hey, let's compare a 14MP foveon image to a 14MP bayer image!".

I am not even really saying "Hey, let's compare a 4.66x3 with a bayer image that has 12MP as output" which is actually what YOU should be saying. Except of course some reports say the real resolution of the 5D is somewhat less than the D200:

http://forum.shutterbug.com/forum/showflat.php?Cat=0&Number=6171&an=0&page=2

This goes hand in hand with what I keep harping on, that without knowing the exact strength of any given cameras AA filter you cannot really say how much detail is captured. I would bet that the 5D has a stronger AA filter, after all Canon is known to prefer that.

What in fact I DID say I thought there would be somewhat less detail but also fewer artifacts. What is a little hard to say is how strong the AA filter on the 5D is which would degrade the effective detail it captured and let the Foveon pull closer...

Honestly I'm not sure why you guys are jumping on the "need to compare a foveon image". My original post was just declaring that the 5D crop was not the beacon of perfection the poster seemed to think it was (look at hw they posted it below a second time as a proud example of what they bayer camera could do for sharpness), nor quite as sharp as it really appears to be. It was just to point out very real flaws in the image, and your response smacks or trying to deflect the argument to ground you feel you can "win", a fruitless endevour on your part if history is any judge. Well I'm having none of your misdirective posting, let's talk artifacts if you want to talk again.

Are you denying there is a very ugly maze pattern in the window on the left that should not be there?

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmasd9/user_home
 
LMC54 said:
All this shows me is that you are the worst in processing and
interpolating, interpolating from a jpeg is really nonsense. Give
me the actual RAW and I'll put that 10meg Bayer image to shame.
Click to expand...
This has been done to death ad infinitem. The results are always similar to this. If you want the Raw you will have to ask Phil, the samples are from his review. But there is no way the sigma image would do so from the raw as the spatial data, just isn't comparable. 6MP bayer yes, 8MP bayer at a stretch, but 10MP bayer is just too large a reach.
 
Whoah, there...
Let's stop and get back on the track.

I never said that picture doesn't exhibit typical "bayerian" artefacts or that it doesn't suffer from moire and aliasing (although aliasing happens on Foveon too, but that's a wholly different subject). I never said it's pixel-perfect either.

What i said was - it has almost as much spatial resolution AS ADVERTISED. Unlike what people here keep suggesting, trying to justify Sigma's-magical-new-physics-defying-interpolated-mode that will once and for all wipe the floor with all the Bayer sensors out there from 10 megapixels up.

So, yes, that photo is a typical bayer interpolation with all it's weaknesses and artifacts, but it doesn't show any sort of extreme loss of detail people keep bringing up and it certainly isn't equivalent to a 4MP Foveon in terms of luma resolution.

On a side note, i'll try to dig up the original RAW and convert it into a completely unsharpened JPG and TIFF.

--
--------------------------------------------
Ante Vukorepa
 
Peter G said:
LMC54 said:
All this shows me is that you are the worst in processing and
interpolating, interpolating from a jpeg is really nonsense. Give
me the actual RAW and I'll put that 10meg Bayer image to shame.
Click to expand...
This has been done to death ad infinitem. The results are always
similar to this. If you want the Raw you will have to ask Phil, the
samples are from his review. But there is no way the sigma image
would do so from the raw as the spatial data, just isn't
comparable. 6MP bayer yes, 8MP bayer at a stretch, but 10MP bayer
is just too large a reach.
Click to expand...
wanna make a wager?
Larry

--
http://www.fredmiranda.com/hosting/showgallery.php?ppuser=235&cat=500
http://www.pbase.com/lmc54/sd10
 
You need to have a source with enough spatial detail to start with to see any difference between DCRAW's AHD and other "more conventional" converters. If your image doesn't contain the spatial information, then there certainly is no method able to "invent" it and put it back there, is it :)

When i dig the original up i'll post a side-by-side comparing a DCRAW conversion with ACR if that's ok with you.
Peter G said:
I am no zeolot for either side. I use Bayer and I have tried many
converters, RSE, Silkypix, ACR, RIT and DCRAW are currently
installed. I have found no resolution advantage with the G6 from
DCRAW. If you could provide a RAW that demonstrates this, I would
like to be enlightened.
Click to expand...
--
--------------------------------------------
Ante Vukorepa
 
Ante Vukorepa said:
You need to have a source with enough spatial detail to start with
to see any difference between DCRAW's AHD and other "more
conventional" converters. If your image doesn't contain the spatial
information, then there certainly is no method able to "invent" it
and put it back there, is it :)

When i dig the original up i'll post a side-by-side comparing a
DCRAW conversion with ACR if that's ok with you.
Click to expand...
When I tested recent DCRAW (using verbose to see that it was AHD demosaic), I got more detail and less noise from RSE.
 
Ante Vukorepa said:
Again, i'm not trying to say Bayer has no artifacts, that it has no
aliasing, that it has a better or equal per-pixel color accuracy.
I'm just saying that a modern interpolation from a 13MP Bayer
reveals much more spatial information that people here are giving
it credit for. Which means it's plain silly to expect a 14MP
interpolation from an SD14 to be of equal spatial resolution as a
13MP Bayer. 8MP, probably. 10 with quite a stretch. But for god's
sake, stop expecting miracles from a "plain ole" marketing trick.
Click to expand...
Fine lets talk only spatial resolution. Over and over again in this very thread the green pixel count has been mentioned, which on the Foven is 4.66 MP and on a 10MP bayer camera is 5MP. On a 13MP sensor the count is more like 6.5.

I don't think it's the perfect judge but I do think it gives you a good lower limit on what the Fovoen capabiltiies are compared to a "perfect" bayer sensor - rather that being "probably 8" with a cap at 10, it seems far more realistic to assume that in fact you get "probably" 9.5 with a cap upwards of 11 or so, a range you have to give because bayer images do not capture all detail perfectly (as your crop has shown).

So the 5D crop would probably have somewhat more detail, but not all that much more with an effective output of 12MP (which is what you get after processing)

But wait - do you REALLY get all of those extra delightful green photosites to use for detail? No you do not, because you seem to have forgotten that they sit behind a veil of secrecy. Ok, not secrecy but a veil notheless - it's called an AA filter, and what is secret is how thick of a veil you are getting with any given camera.

For example, I really hate to drag other peoples arguments into this forum but some people seem to consider the D200 as actually having greater resolvement powers than the 5D:

http://forum.shutterbug.com/forum/showflat.php?Cat=0&Number=6171&an=0&page=2

If this is true you can thank that veil. Even if it's not quite true you have to factor in the mystery of the veil when saying how much detail a camera can really resolve, which makes any atempt to imagine just how close an SD-14 crop would look to that 5D crop an excersize in futility.

It is my personal belief that with greater resolution counts camera makers have also increased the strength of the AA filters, because the greater spatial resolution you mentioned lets them use a stronger veil without anyone being able to tell. That is kind of the opposite conclusion you are going for (that modern bayer sensors are resolving more detail), but I would say evidence from comparing bayer cameras alone points rather to increase in raw MP count also leading to stronger AA filters.

As a side note this is eactly my problem with MP as a measurement of anything, if D200 and 5D owners can get into such a fray what meaning does that number really have?

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmasd9/user_home
 
LMC54 said:
wanna make a wager?
Larry
Click to expand...
Absolutely.

I am sure that if you had Phils source Raw you couldn't best the 10MP bayer output.

If you can find another fairly shot image (fairly framed, focused and no red/blue res charts), that the result will be the same.

I have seen the results from the time the first Foveon samples were released. I know what each technology delivers.
 
You must log in or register to reply here.

Keyboard shortcuts

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