Mega Pixel Myth

On May 23, 10:38 pm, acl wrote:

Anyway, I'll print a photo at 100 and 300ppi, photograph it and post
it later to see what the difference is (or isn't, I accept that I may
be imagining that there is a difference in my prints).

I took this shot
http://www.pbase.com/al599/image/79058890
and printed a part of it at 480ppi, 300ppi and 100ppi. I started by
downsampling it to some particular size and 480ppi; I printed that,
then, in photoshop, downsampled by selecting bicubic resampling and
decreasing the resolution to 300, printed that, went back to the
480ppi image, downsampled to 100ppi in the same way, and printed
again. I then photographed the results at 1:2 (or closer to 1:1.9 in
fact), cropped a part, changed to srgb, and put them he
http://www.pbase.com/al599/image/79296092/original
http://www.pbase.com/al599/image/79296094/original
http://www.pbase.com/al599/image/79296096/original
As you can see, there is a huge difference between the 100ppi and
300ppi prints (and it looks even worse in reality, believe me). The
problem is not the bicubic downsampling, on screen it looks fine (ie
no more jaggies than you'd expect). The 480ppi has some more details
(there's a fence in the crop, check that) but in reality it's not much
(but it does look crisper). I think it's a result of some
interpolation by the driver, as on screen the fence is clearly
resolved at both 300ppi and 480ppi, while on the prints the 300ppi one
is not so clearly resolved.

Anyway, if you think before posting this time, you'll see that at a
sensor length of 23.5mm (call it 1 inch) and 1:2 magnification, you'd
expect to see something like 200 image pixels along the long
dimension. That is around what I see (of course the image I linked to
is a crop, not the full frame, although I can put that there too).

So maybe we can now agree that you were wrong, and 300ppi does look
significantly better than 100ppi.

In reality, as I said, the 100ppi print looks extremely bad;
personally, I wouldn't bother with printing if that was the maximum
quality. After all, it's like a monitor. Anyway, that you claim that
100ppi is as sharp as 300ppi for inkjets tells me that either you have
never seen 100ppi prints (so you don't know how it looks like), never
seen inkjet prints (so you think they look like 100ppi prints), or
both.

Anyway. I am curious to see what you will now think of to avoid saying
you're wrong.

"acl" wrote in message
oups.com...

Anyway. I am curious to see what you will now think of to avoid saying
you're wrong.


That's easy.. note I said most inkjets.
I know there are some high resolution multiple colour inkjets with variable
drop sizes that will do better than 100 ppi but most of them being used
don't.
If you compared your 300ppi print to a 300ppi dyesub print you would see a
bigger difference.

Btw can you see the difference when you view the prints rather than
enlarging them.
100 ppi is better than many magazines use and better than most monitors.

On May 24, 12:46 am, "dennis@home"
wrote:
"acl" wrote in message

oups.com...

Anyway. I am curious to see what you will now think of to avoid saying
you're wrong.

That's easy.. note I said most inkjets.

Yes, I suspected that you'd make me cut and paste what you wrote to
prove I am not inventing stuff. In fact, you said
The same print done on an inkjet would only need about 100 ppi so about 0.25
MP (maybe 200 ppi @ 1 MP on an expensive inkjet).
and
On an inkjet you have to be able to print ~750 dots to represent a pixel
(256x3). This means the ppi is about twelve-sixteen times less than the dpi
for the printer.
which I imagine is where you got 100ppi, you divided 1440 (or so) by
around 12 to get around 120ppi. Basically, you repeatedly made the
point that you need many dots to make a pixel in an inkjet (which is
of course true). Obviously the reason you are wrong is related to
variable drop sizes and multiple inks, but that is not what you said,
is it?
I know there are some high resolution multiple colour inkjets with variable
drop sizes that will do better than 100 ppi but most of them being used
don't.
I've seen prints from a 80 euro epson printer (not so expensive), and,
besides it using dyes instead of pigments like the R800 (which doesn't
seem to affect image quality either way), I could not see much
difference (none whatsoever with the naked eye). So it's not the
price.

And if you know it, why did you keep insisting that 100ppi is all you
need? And when I showed you the photos that indicated that this isn't
true, you kept arguing (the contrast is high, the lines are not
diagonal etc). I hate going on about this, but at least don't treat
people as if they are idiots: once you couldn't continue to claim that
what I said isn't true, you switched to "but that's not what I was
saying". Right.

If you compared your 300ppi print to a 300ppi dyesub print you would see a
bigger difference.

Well great, your dye-sub printer is better. Excellent. What does this
have to do with this thread?

Btw can you see the difference when you view the prints rather than
enlarging them.

Did you bother reading my posts? I said it three times in two separate
posts.

100 ppi is better than many magazines use and better than most monitors.

And yet it looks bad on print. There was a thread about this here,
earlier. People said the usual things there (what you said,
basically), and I again got no answer when I showed these photos. I
don't know. Come to think of it, you probably were in that thread too.
And I vaguely remember that you also at some point started telling
people that DOF does not depend on the print size. Hmm... Don't
remember if you did admit that you were wrong then. Suspect not

Oh well. I think I've had enough. I don't like this kind of thing, but
this is ridiculous. If you don't know what you are talking about,
don't write as if you're an expert, and certainly don't insist.

"acl" wrote in message
oups.com...

I've seen prints from a 80 euro epson printer (not so expensive), and,
besides it using dyes instead of pigments like the R800 (which doesn't
seem to affect image quality either way), I could not see much
difference (none whatsoever with the naked eye). So it's not the
price.

Ah well there is a difference.. try overlaying the dots on them.
Leave them in the sun for a week.

And if you know it, why did you keep insisting that 100ppi is all you
need?

Because it is for most people.
Its one hundreth of an inch on the print.. most people can't see much
better.
If you like to argue about print quality then maybe it isn't.
But if you need that much quality inkjets are not what you want.

And when I showed you the photos that indicated that this isn't
true, you kept arguing (the contrast is high, the lines are not
diagonal etc). I hate going on about this, but at least don't treat
people as if they are idiots: once you couldn't continue to claim that
what I said isn't true, you switched to "but that's not what I was
saying". Right.

If you compared your 300ppi print to a 300ppi dyesub print you would see
a
bigger difference.

Well great, your dye-sub printer is better. Excellent. What does this
have to do with this thread?

Btw can you see the difference when you view the prints rather than
enlarging them.

Did you bother reading my posts? I said it three times in two separate
posts.

100 ppi is better than many magazines use and better than most monitors.

And yet it looks bad on print.

It looks bad on your prints.. that doesn't mean it looks bad on all prints.

There was a thread about this here,
earlier. People said the usual things there (what you said,
basically), and I again got no answer when I showed these photos. I
don't know. Come to think of it, you probably were in that thread too.
And I vaguely remember that you also at some point started telling
people that DOF does not depend on the print size. Hmm... Don't
remember if you did admit that you were wrong then. Suspect not

It doesn't.
DoF is in the way the image is recorded not in how it is displayed.
If it depended on print size it would also depend on viewing distance.


Oh well. I think I've had enough. I don't like this kind of thing, but
this is ridiculous. If you don't know what you are talking about,
don't write as if you're an expert, and certainly don't insist.


Well you aren't in a position to judge who is an expert or not based on what
you have said here.

On May 24, 3:49 am, "dennis@home"
wrote:
"acl" wrote in message

And yet it looks bad on print.

It looks bad on your prints.. that doesn't mean it looks bad on all prints.

See, that's what I thought would happen. I showed you what it looks
like (enlarged), you keep saying it doesn't look bad. Well I can't
show an actual print here, sorry. It looks terrible from arm's length.
You can disagree all you want, but anybody who has a printer can
easily try it and see for themselves. I can't really stop you from
saying "no it doesn't", so I give up.

And I vaguely remember that you also at some point started telling
people that DOF does not depend on the print size. Hmm... Don't
remember if you did admit that you were wrong then. Suspect not

It doesn't.
DoF is in the way the image is recorded not in how it is displayed.
If it depended on print size it would also depend on viewing distance.

Well ok, keep insisting. You could google "depth of field" and see how
many places discussing depth of field agree with you and how many do
not. Of course, weight of numbers does not ensure truth, but this
question is simple enough that a few minutes of thought would convince
most people, so numbers is a good indicator.



Oh well. I think I've had enough. I don't like this kind of thing, but
this is ridiculous. If you don't know what you are talking about,
don't write as if you're an expert, and certainly don't insist.

Well you aren't in a position to judge who is an expert or not based on what
you have said here.

Luckily, we have an audience

"acl" wrote in message
ups.com...
On May 24, 3:49 am, "dennis@home"
wrote:
"acl" wrote in message

And yet it looks bad on print.

It looks bad on your prints.. that doesn't mean it looks bad on all
prints.

See, that's what I thought would happen. I showed you what it looks
like (enlarged), you keep saying it doesn't look bad. Well I can't
show an actual print here, sorry. It looks terrible from arm's length.
You can disagree all you want, but anybody who has a printer can
easily try it and see for themselves. I can't really stop you from
saying "no it doesn't", so I give up.

Why give up?
As with most photography it is subjective.
You can be right and I can be right.
It is not an absolute.


And I vaguely remember that you also at some point started telling
people that DOF does not depend on the print size. Hmm... Don't
remember if you did admit that you were wrong then. Suspect not

It doesn't.
DoF is in the way the image is recorded not in how it is displayed.
If it depended on print size it would also depend on viewing distance.

Well ok, keep insisting. You could google "depth of field" and see how
many places discussing depth of field agree with you and how many do
not. Of course, weight of numbers does not ensure truth, but this
question is simple enough that a few minutes of thought would convince
most people, so numbers is a good indicator.

Look at it like this..

If I take a picture the image will not get any sharper once the cone of
light from the lens is smaller than the resolution of the sensor/film.
Anything that intersects the sensor/film with a circle smaller than the
resolution will be as well focused as the medium can record.


Now you say it depends on the print so I take a 35mm negative and blow it up
to 50 inches.
Now view it from 10 inches and its all fuzzy but you can still see which
bits are in focus and which aren't.

Now view it from 20 feet and it looks much sharper but you can still see
which bits are in focus and which aren't.

The DoF has not changed because of the printing.
Nothing you do in the printing will bring out of focus detail into focus.


Now you could print it at 1:1 and then it may all appear to be sharp but
that doesn't mean that the DoF has changed just that you can't see it.


Oh well. I think I've had enough. I don't like this kind of thing, but
this is ridiculous. If you don't know what you are talking about,
don't write as if you're an expert, and certainly don't insist.

Well you aren't in a position to judge who is an expert or not based on
what
you have said here.

Luckily, we have an audience


I haven't seen any vote yet.
I don't see why you feel the need to argue about a subjective think like
print quality on inkjets the way you do.
However for the DoF argument you are just plain wrong as it is a
characteristic of the recording system and not the reproduction system
(although they may have DoF issues of their own but that is a design
problem).

dennis@home offered these thoughts for the group's consideration
of the matter at hand:

See, that's what I thought would happen. I showed you what it
looks like (enlarged), you keep saying it doesn't look bad.
Well I can't show an actual print here, sorry. It looks
terrible from arm's length. You can disagree all you want,
but anybody who has a printer can easily try it and see for
themselves. I can't really stop you from saying "no it
doesn't", so I give up.

Why give up?
As with most photography it is subjective.
You can be right and I can be right.
It is not an absolute.

Couldn't agree more! Fitness of purpose is - or should be - the
main criteria and what is important to portrait photographers vs.
wildlife vs. scenery vs. sports vs. architectural, even my fav
subject, cars, what constitutes "good" in a camera or a printer
is far too qualitative and subjective to make strong assertions
about. Easiest example I can think of is portraiture, where
photographers often WANT a soft image so that their "subjects"
think they look better than they really do, they use a mild
telephoto because it de-emphasizes large noses and ears, etc.
And, people often want their portraits on matt paper so they
don't have to buy an expensive frame and glass for photos they
take of their own family. OTOH, people who prize fine detail
usually want glossy paper because it both perceptually and in
reality better for that.

Maybe that's why Baskin-Robbins has so many flavors of ice cream?

Well ok, keep insisting. You could google "depth of field"
and see how many places discussing depth of field agree with
you and how many do not. Of course, weight of numbers does
not ensure truth, but this question is simple enough that a
few minutes of thought would convince most people, so numbers
is a good indicator.

Look at it like this..

If I take a picture the image will not get any sharper once
the cone of light from the lens is smaller than the resolution
of the sensor/film. Anything that intersects the sensor/film
with a circle smaller than the resolution will be as well
focused as the medium can record.

As I try to recall my early days as a 35mm photographer, I
learned this strange term "circle of confusion" upon which a
standard was created to be able to predict DOF. Mathematically,
there is only one focus setting that is truly sharp, the rest is
tricking the eye.

Mathematicians will tell you that it is impossible to prove a
hypothesis by citing examples, ala Google, because all it takes
is one exception to disprove everything. It is equally impossible
to try to prove a negative or null hypothesis for the same
reason. But, if the Google hits are describing mathematically
correct methods for any of these discussion items du jour, then,
yes, citing "examples" does work.

Now you say it depends on the print so I take a 35mm negative
and blow it up to 50 inches.
Now view it from 10 inches and its all fuzzy but you can still
see which bits are in focus and which aren't.

Now view it from 20 feet and it looks much sharper but you can
still see which bits are in focus and which aren't.

The DoF has not changed because of the printing.
Nothing you do in the printing will bring out of focus detail
into focus.

Again, I highly agree. The same basic argument you make can also
apply to the never ending debate about how many MP and how much
PPI is really needed. My first question after I find out what
size people want to print at is "how far do you intend to view
these from?"

Now you could print it at 1:1 and then it may all appear to be
sharp but that doesn't mean that the DoF has changed just that
you can't see it.

I haven't seen any vote yet.
I don't see why you feel the need to argue about a subjective
think like print quality on inkjets the way you do.
However for the DoF argument you are just plain wrong as it is
a characteristic of the recording system and not the
reproduction system (although they may have DoF issues of
their own but that is a design problem).

At least there is sound mathematics behind DOF. There are
obviously agreed upon norms to judge print quality but your point
is still well taken. Can't imagine a repro system having a DOF
issuee unless it is using some sort of optical projector. Nobody
ever argued about DOF when creating the old-fashioned chemical
prints; when focusing the enlarger, the "image" on the paper
either is or is not in focus.

--
HP, aka Jerry

On 23 May 2007 13:17:54 -0700, acl wrote:

I took this shot
http://www.pbase.com/al599/image/79058890
and printed a part of it at 480ppi, 300ppi and 100ppi. I started by
downsampling it to some particular size and 480ppi; I printed that,
then, in photoshop, downsampled by selecting bicubic resampling and
decreasing the resolution to 300, printed that, went back to the
480ppi image, downsampled to 100ppi in the same way, and printed
again. I then photographed the results at 1:2 (or closer to 1:1.9 in
fact), cropped a part, changed to srgb, and put them he
http://www.pbase.com/al599/image/79296092/original
http://www.pbase.com/al599/image/79296094/original
http://www.pbase.com/al599/image/79296096/original
As you can see, there is a huge difference between the 100ppi and
300ppi prints (and it looks even worse in reality, believe me). The
problem is not the bicubic downsampling, on screen it looks fine (ie
no more jaggies than you'd expect).

Did you send 100 ppi image directly to the printer? If so I'm almost
positive that the problem was in the primitive upsampling algorithm your
printer uses, most likely nearest neigbour. Every ink jet printer resamples
the image to it's internal resolution before starting to render it by the
halftoning algorithm. For home epsons that's 720 ppi.

So the squares you got are likely the result of printer upscaling the image
from 100 ppi to 720 ppi using some primitive algorithm.

I took your original image and reduced the resolution 7.2 times using the
bicubic algorithm. Then I upscaled it back to the original size, first
using the nearest neighbour and then bicubic algorithm. Here is what I got:

Nearest neighbour:
http://img248.imageshack.us/img248/3...eighborux6.jpg

Bicubic:
http://img126.imageshack.us/img126/696/bicubicmr4.jpg

Looks a bit like the difference between your photos, does it not?

To do a proper test you should upscale 480, 300 and 100 ppi images to 720
ppi in photoshop and apply the appropriate amount of sharpening for each
image. I bet that the difference won't be nearly as big.

dennis@home wrote:
"acl" wrote in message
oups.com...

I've seen prints from a 80 euro epson printer (not so expensive), and,
besides it using dyes instead of pigments like the R800 (which doesn't
seem to affect image quality either way), I could not see much
difference (none whatsoever with the naked eye). So it's not the
price.

Ah well there is a difference.. try overlaying the dots on them.
Leave them in the sun for a week.

And if you know it, why did you keep insisting that 100ppi is all you
need?

Because it is for most people.
Its one hundreth of an inch on the print.. most people can't see much
better.
If you like to argue about print quality then maybe it isn't.
But if you need that much quality inkjets are not what you want.

snip

This is really too funny.
I use an Epson R1800, an A3 printer but very similar to the Epson R800.
You should read this first:
http://www.crystalcanyons.net/pages/...00Printer.shtm
Then also consider that the people using the R800 for high resolution
card printing of "stereo cards" don't know (or didn't know at the time)
that resampling of the image prior to printing to a divisor of the
native resolution does have a noticeable impact on the printed
resolution that can be achieved. "Normally" that will be 360ppi for
Epson, and 300 ppi for Canon / HP. The benefit of resampling above that
can be seen, but practical benefit for normal photographs is pretty
doubtful. 300 or 360 dpi (from 300 or 360 good pixels per inch) is
about as good as anyone would normally want for small photographs. 100
dpi is not enough, but YMMV. You can save yourself the bother of
resampling by using QImage software to do it automatically. Compared to
Fuji Frontier / Crystal Archive light jet prints (300dpi), the Epson
prints with 360dpi look at least as sharp and detailed, possibly more
so. Dot pattern is not visible to the naked eye. With a loupe, dot
pattern is visible, and appears very similar to film grain on frontier
prints from scanned iso 100 colour negative, printed at 6x4, and I don't
believe that grain has ever been considered an issue with 35mm iso 100
printed that small.
Under my microscope at ~50x, you can see that the ink droplets are quite
defined - not "splattered" out. You can also see the effect of the part
gimmick "RPM" optimised 5760 dpi mode (available with some desktop Epson
printers) on drop placement. The RPM mode makes very little difference
to the naked eye, except I get the impression that shadow detail seems
somewhat improved.
Anyone disputing the quality of inkjet prints these days must have been
asleep for a few years.
OTOH, the inkjet makers are fleecing everyone with exorbitant pricing.
Canon and HP are now serious players in the inkjet game, so eventually
there is hope that there will be some price competition.

On May 24, 1:16 pm, DeanK wrote:
On 23 May 2007 13:17:54 -0700, acl wrote:



I took this shot
http://www.pbase.com/al599/image/79058890
and printed a part of it at 480ppi, 300ppi and 100ppi. I started by
downsampling it to some particular size and 480ppi; I printed that,
then, in photoshop, downsampled by selecting bicubic resampling and
decreasing the resolution to 300, printed that, went back to the
480ppi image, downsampled to 100ppi in the same way, and printed
again. I then photographed the results at 1:2 (or closer to 1:1.9 in
fact), cropped a part, changed to srgb, and put them he
http://www.pbase.com/al599/image/79296092/original
http://www.pbase.com/al599/image/79296094/original
http://www.pbase.com/al599/image/79296096/original
As you can see, there is a huge difference between the 100ppi and
300ppi prints (and it looks even worse in reality, believe me). The
problem is not the bicubic downsampling, on screen it looks fine (ie
no more jaggies than you'd expect).

Did you send 100 ppi image directly to the printer? If so I'm almost
positive that the problem was in the primitive upsampling algorithm your
printer uses, most likely nearest neigbour. Every ink jet printer resamples
the image to it's internal resolution before starting to render it by the
halftoning algorithm. For home epsons that's 720 ppi.

So the squares you got are likely the result of printer upscaling the image
from 100 ppi to 720 ppi using some primitive algorithm.

Yes of course it looks like nearest neighbour. The idea is that there
are things that are well resolved in the 300ppi print that simply do
not exist in the 100ppi. These won't be created by more sophisticated
upsampling, that will just smooth jagged lines. Also, the mere fact
that you can clearly see the jaggies, and they're well-defined, means
that the printer can resolve a lot more than the pixels, which is what
I am arguing.

Also, I could have downsampled more carefully (one could argue). But I
didn't. I just wanted to show that 100ppi is hardly the upper limit of
what can be resolved.


I took your original image and reduced the resolution 7.2 times using the
bicubic algorithm. Then I upscaled it back to the original size, first
using the nearest neighbour and then bicubic algorithm. Here is what I got:

Nearest neighbour:http://img248.imageshack.us/img248/3...eighborux6.jpg

Bicubic:http://img126.imageshack.us/img126/696/bicubicmr4.jpg

Looks a bit like the difference between your photos, does it not?


Right. I was trying to show that more than 100ppi is resolved by the
printer. I didn't want to compare upsampling algorithms, I just get
irritated when people authoritatively state rubbish like "inkjet
printers cannot resolve more than 100ppi" or "3MP are enough for A4
prints". They may be, depending on the subject, the material on which
you are printing etc, but as a general statement...

To do a proper test you should upscale 480, 300 and 100 ppi images to 720
ppi in photoshop and apply the appropriate amount of sharpening for each
image. I bet that the difference won't be nearly as big.

It depends what you want to test. I did not want to test what the
maximum quality is for 100ppi, but whether or not 300ppi makes any
difference. And in fact I have had to print something at 150ppi a few
times, no matter how much effort I spent upsampling and carefully
masking and sharpening etc, it still looked rubbish (it was a
reasonably detailed scene though; a moody portrait may be fine down to
100ppi or even less, sometimes).