Resolution and print size

"Marc Wossner" wrote:
On 25 Jan., 15:29, "David J. Littleboy" wrote:
"Marc Wossner" wrote:

This guy complains about moiré and fine text in street signs getting
messed
up, but it sure looks to me that the 5D and 645 (actually, 6x6 cropped to
slightly smaller than 645) are pretty much equivalent for prints 17x26"
and
smaller. http://www.shortwork.net/equip/review-1Ds-SQ-scantech/

So there´s the question again: How can you print that large from the
resolution the 5D gives?


Well, don't ask me; I think printing 645 larger than 13x19 is nuts. But 35mm
folks make 16x20s all the time. (I saw an exhibit of 20x30s at the Nikon
gallery here that was incredibly soft mush; they looked bad from across the
room.)

Seriously, though, if you have a decent image, the larger you print it, the
better it looks. Even if the result is mush. So no one has ever been unhappy
with a large print they've made from their own images.


When you calculate 4368 pixels * 0,7 = 3057,6 lines / 2 = 1529 line
pairs / 35,8 mm you reach 42,7 lp/mm.
I don´t know the true resolution of the lens, but as it´s a prime
lens let´s assume 100 lp/mm.
Taking it into account like this 1/T = 1/43 + 1/100 makes it a system
resolution of 30 lp/mm worst case.


Let's see. The 5D produces 2912 x 4374 pixels (plus or minus a few depending
on the RAW converter), and digital images tend to have a practical
resolution of about one line pair per 3 pixels. So that's 970 lp/in or 38
lp/mm. OK. Our numbers agreeg.

In real life, the 5D produces about 240 ppi at 12x18 (1/2" borders on 13x19
paper). The prints look gorgeous. That's a 12x enlargement with 3 lp/mm
resolution. That's pretty cool, because film look really really bad at a 12x
enlargement. Either grainy or mushy or both.


The images presented are printed at 17x26" so that´s a magnification
of x18.
Divide the 30 lp/mm by the 2 lp/mm that are necessary for that diagonal
you get a max magnification of x15.
But the guy shows clippings that appear sharp from 10" and for that
distance you need 6,88 lp/mm to have a sharp impression.


My theory on what's going on here is that your "you need 6,88 lp/mm to have
a sharp impression" is due to misinformation from the dizzy Leica folks.

What they do is shoot high contrast (1:1000) test charts and assume that
their real life images achieve similar levels of detail capture; they think
that tests with strobe illuminated charts with the camera bolted to granite
test bench somehow relates to handheld street shooting. My experience
shooting film (and the data sheets from the mfrs) is that you see nothing of
the sort in real life images _other than street signs_. Film types (and
I've made this mistake myself) will look at a street sign image and think
how cool it is that their camera resolves it, while failing to notice that
the textures at that enlargement not only aren't captured, but are
completely swamped by the grain noise.


Note that those figures are all based on "normal" 20/20 vision or max
resolvable detail of 1 minute of an arc (see Norman Koren´s site for
the calculation: http://www.normankoren.com/Tutorials/MTF.html).
As I know that you are quite knowledgeable from various other threads
David, can you explain me how this is possible, or better, where my
possible misconception is?


He's looking at film images that are grossly soft as well. I've printed a
lot of test images at that magnification from film, and they're mush.

David J. Littleboy
Tokyo, Japan

In real life, the 5D produces about 240 ppi at 12x18 (1/2"
borders on 13x19 paper). The prints look gorgeous. That's a
12x enlargement with 3 lp/mm resolution. That's pretty cool,
because film look really really bad at a 12x enlargement.
Either grainy or mushy or both.

That is one of the great things about digital - no grain.

My 5mp Oly C5050 does 8x10 at 240 ppi and it looks at least as
good as 8x10's that I have taken and had printed with my 35mm
film camera. Sometimes it is obviously clearer even if the level
of detail is not as great overall.

On 26 Jan., 01:42, "David J. Littleboy" wrote:
"Marc Wossner" wrote:
On 25 Jan., 15:29, "David J. Littleboy" wrote:
"Marc Wossner" wrote:

This guy complains about moiré and fine text in street signs getting
messed
up, but it sure looks to me that the 5D and 645 (actually, 6x6 cropped to
slightly smaller than 645) are pretty much equivalent for prints 17x26"
and
smaller. http://www.shortwork.net/equip/revie...SQ-scantech/So there´s the question again: How can you print that large from the
resolution the 5D gives?


Well, don't ask me; I think printing 645 larger than 13x19 is nuts. But 35mm
folks make 16x20s all the time. (I saw an exhibit of 20x30s at the Nikon
gallery here that was incredibly soft mush; they looked bad from across the
room.)

Seriously, though, if you have a decent image, the larger you print it, the
better it looks. Even if the result is mush. So no one has ever been unhappy
with a large print they've made from their own images.

When you calculate 4368 pixels * 0,7 = 3057,6 lines / 2 = 1529 line
pairs / 35,8 mm you reach 42,7 lp/mm.
I don´t know the true resolution of the lens, but as it´s a prime
lens let´s assume 100 lp/mm.
Taking it into account like this 1/T = 1/43 + 1/100 makes it a system
resolution of 30 lp/mm worst case.


Let's see. The 5D produces 2912 x 4374 pixels (plus or minus a few depending
on the RAW converter), and digital images tend to have a practical
resolution of about one line pair per 3 pixels. So that's 970 lp/in or 38
lp/mm. OK. Our numbers agreeg.

In real life, the 5D produces about 240 ppi at 12x18 (1/2" borders on 13x19
paper). The prints look gorgeous. That's a 12x enlargement with 3 lp/mm
resolution. That's pretty cool, because film look really really bad at a 12x
enlargement. Either grainy or mushy or both.

The images presented are printed at 17x26" so that´s a magnification
of x18.
Divide the 30 lp/mm by the 2 lp/mm that are necessary for that diagonal
you get a max magnification of x15.
But the guy shows clippings that appear sharp from 10" and for that
distance you need 6,88 lp/mm to have a sharp impression.


My theory on what's going on here is that your "you need 6,88 lp/mm to have
a sharp impression" is due to misinformation from the dizzy Leica folks.

What they do is shoot high contrast (1:1000) test charts and assume that
their real life images achieve similar levels of detail capture; they think
that tests with strobe illuminated charts with the camera bolted to granite
test bench somehow relates to handheld street shooting. My experience
shooting film (and the data sheets from the mfrs) is that you see nothing of
the sort in real life images _other than street signs_. Film types (and
I've made this mistake myself) will look at a street sign image and think
how cool it is that their camera resolves it, while failing to notice that
the textures at that enlargement not only aren't captured, but are
completely swamped by the grain noise.

Note that those figures are all based on "normal" 20/20 vision or max
resolvable detail of 1 minute of an arc (see Norman Koren´s site for
the calculation:http://www.normankoren.com/Tutorials/MTF.html).
As I know that you are quite knowledgeable from various other threads
David, can you explain me how this is possible, or better, where my
possible misconception is?


He's looking at film images that are grossly soft as well. I've printed a
lot of test images at that magnification from film, and they're mush.

No. I do analog photography for a long time and started studying the
perception side to make better use of the medium. That´s where all I
know is derived from. The 6,88 lp/mm are calculated using only figures
of our visual system. Koren´s site explains this and all the
informations are backed up by a lot of scientifc research. 1 minute of
an arc *is* the resolution limit of the *average* human and the 6,88 is
based on that. But you can also reach it from another side:

cycles/degree = 600/Snellen denominator = 600/20 = 30 cycles/degree
cycles/degree * (180/pi) * (1/distance in mm) = 30 * (180/pi) * (1/250
mm) = 6,87 lp/mm

As stated this is an average value and a lot of people can see much
better, in the range of 60 cycles/degree.
That´s what I know and where it comes from. It holds true in analog
photography and that´s why I don´t understand what´s going on in the
digital case.

Marc

Forgive my selective editing, David, but in these three bits, I think
you summed it up well.

On Jan 26, 10:42 am, "David J. Littleboy" wrote:
Seriously, though, if you have a decent image, the larger you print it, the
better it looks. Even if the result is mush...

My theory on what's going on here is that your "you need 6,88 lp/mm to have
a sharp impression" is due to misinformation...

What they do is shoot high contrast (1:1000) test charts and assume that
their real life images achieve similar levels of detail capture...

Exactly. The 300 ppi thing (which is derived from the 6.88) is almost
useless, imo, as a guide to enlargability, unless you are only
enlarging resolution test charts. If the image content is any of the
following, then you can decrease that figure, sometimes dramatically:

- close up macros without fine detail (eg flower macros, not insects
with lots of tiny hairs..)
- very tightly cropped portraits (because getting too close to a face
feels uncomfortable to the viewer)
- soft focus portraits (obviously)

However, if your image is:
- a landscape (is that grass/foliage/pine needles.. or mush(â„¢David)?)
- cityscape (is that blurry bit my house, or why can't I read that
little sign?),
- beachscape (is that girl way back there topless?)
- or a large group portrait (is that Uncle Ed or Aunt Martha, 3rd row,
10th from the end?)

I proved this to myself long ago when I first tested out the much
maligned Sony DSCF828 (which does produce very nice and quite sharp 8Mp
images.) I blew a series of images up to 13" x 19" prints (ie about
190 ppi), including portraits like this one:

http://www.marktphoto.com/portrait/s...ppiness_is.jpg
(sorry, slightly oversharpened - you'll have to trust me that the
original is nailed!)

and also landscapes and other images that invited close inspection,
like this:

http://www.marktphoto.com/marina_details.jpg

The portraits looked stunning, and yet the detailed shots like the
second example looked mediocre. You would have sworn they were taken
on two different cameras. But on close inspection, you could see that
the portrait was indeed exactly as sharp as the other - it was a
combination of the way you viewed the images (the portrait 'made' you
want to view it at about 18" at which point you could see all the
*useful* detail, right down to the detail in the irises and
eyelashes.., but the landscape invited much closer inspection and you
could more easily see the very slight pixellation, even at the same
viewing distance). There's a lot of factors in this, eg diagonal
jaggies are much more likely to show up in boat rigging than eyes...
and of course how well you post process, what method you interpolate
with, etc...

Anyway, if you add it all up, your brain will tell you how sharp the
print is, but it is very likely lying..!

Just use the 300/200/100 thing as a vague guide and experiment..

After digging deeper into this I found a way to explain the facts that
contradict each other at first glance.
The following quote from an old thread lead me the way: "It´s not the
resolution of the image on the retina that makes the difference, it´s
the contrast of the relatively well resolved information." What
matters here is the subjective quality factor (sqf) and the fact that
the better contrast that most digital cameras reproduce at the spatial
frequencys between 5 and 10 cycles/degree, which are most important
for perceived sharpness, allows for larger prints than the resolution
numbers alone predict. The following three websites give further
information:

http://www.bobatkins.com/photography.../mtf/mtf4.html
http://www.imatest.com/docs/sqf.html
http://luminous-landscape.com/forum/...hp/t11480.html

Marc