Most widely available lossless format for documents?

[toby]

No. Pixel is a valid unit. W x H pixels is grand as dimensions go.

It's a count. It's also dimensionless. It's a purely terminological
point. There is no misunderstanding between us, except I was originally
responding to a strange assertion from jjs. Any further debate on this
is a waste of time.

That's true. But "dimension" needn't by "real world."


I wrote nothing above.

I was responding to Hecate.

 

[Kennedy McEwen]

No. Pixel is a valid unit.

I think you mean "Yes. Pixel is a valid unit".
Being dimensionless does not preclude something from being a unit.

 

[Hans-Bernhard Broeker]

All images have physical dimensions.

Not all --- not in the usual meaning of a "physical" dimension.

On a computer they're expressed as pixels.

A pixel is *not* a physical dimension. It's a purely abstract one.
The physical size of an actual pixel is a design parameter of a
computer's display. It differs from machine to machine.

On a print the resolution is expressed as dpi. The dpi is completely
immaterial as far as an image on a computer is concerned.

Not quite. It can be material or not, depending on where the image
currently is. While it's being displayed on the screen, the DPI
resolution of that screen becomes the DPI of the image displayed. If
the image is not visible, it is can be immaterial whether it has an
associated DPI value or not --- but if it does have one, that means
that on displaying it on a physical medium, you have a choice to make:
whether to match pixel counts, or physical size of the image.

E.g. a 600x600 pixel input scanned at 300 DPI is, physically, 2 inches
squared. But a computer's monitor has a lot less than 300 DPI ---
let's say it's 100. So how to display the image: at 600x600 pixels,
which would make it 6 x 6 inches, or at 2x2 inches, downsampling it to
200x200 pixels in the process?

And then there's images that never corresponded to a particular
physical size anyone's interested in --- most prominently, digital
camera shots. The image sensor in a digicam does, of course, have a
physical size, and as such, the image acquired by it has a DPI value
--- but because it's a projected 2D image of 3D real-world objects,
the DPI value of the sensor is quite irrelevant. What you need to
know in this case is the absolute pixel count, i.e. the number of
pixels in the image. The DPI setting is interesting for the people
designing the camera, though --- it governs effects like
susceptibility to electronic noise vs. optical signal strength, and
indirectly, through the physical sensor size, decides how hard the job
is for the optical engineers to make lenses for that camera.

 

[jjs]

It's a count. It's also dimensionless. It's a purely terminological
point. There is no misunderstanding between us, except I was originally
responding to a strange assertion from jjs. Any further debate on this
is a waste of time.

There is no debate. You simply have a silly idea you cannot give up.

 

[jjs]

It's a count. It's also dimensionless.

It is not dimensionless: it represents a value other than zero. If it were
dimensionless then anything made of pixels could be compressed to infinity.

 

[Kennedy McEwen]

It is not dimensionless: it represents a value other than zero. If it were
dimensionless then anything made of pixels could be compressed to infinity.

Rubbish, there are plenty of dimensionless units. Gradient of a slope
is one example most people encounter at an early age. If the slope
increases 1ft in every 10ft then it has a gradient of 0.1ft/ft - the
units cancel out and you have a dimensionless gradient of 10%. It is
pretty clear to everyone learning to walk that a 5% gradient is half as
steep as a 10% one!

Interest rates are another example - for every $10 you invest your bank
will give you $1. Interest rate is 0.1 dollars per dollar, or 10% - no
dimensions. You would be pretty pissed off if your bank then turned
round and told you that they had infinitely compressed their interest
rate and you were just getting back the $10 you invested!

Dimensionless units exist all over the place, just as dimensionless
constants do - pi, e, etc.

 

[jjs]

Rubbish, there are plenty of dimensionless units. [...]
Dimensionless units exist all over the place, just as dimensionless
constants do - pi, e, etc.

You are speaking of metrics. I was just having p(h)un. Have a Guinness on
me.

 

[Tacit]

I never heard of DPI metric stored in an image,

Almost all image formats include the pixel-per-inch information, because
almost all image formats are designed to store images that can be used
for printing, and any image destined for print (or other physical world
output) must include not only the pixels but also some measure of how
large the pixels are.

Take a TIFF file that is 1000 by 3000 pixels. Save it at 300 pixels per
inch, then change the resolution without resampling so that it is 1000
by 3000 pixels and 72 pixels per inch and save it again. Both images
contain absolutely bit-for-bit identical pixel information.

Now place both images in a page layout program such as QuarkXPress. One
will print far larger than the other. Sam epixel information, but the
TIFF file format stores the pixels AND the resolution.

File fo0rmats which do not record resolution information are few and far
between, and are usually formats intended only for on-screen display (eg,
GIF).

...and if it is there,
it's just a handy metric for those who don't understand that pixels-
is-pixels.

It's a metric for those who intend to do more than display an image on a
fixed-pixel device such as a computer monitor. Remember, images don't
just get shown on video displays; sometimes, they get printed, too.

 

[toby]

Rubbish, there are plenty of dimensionless units. ...

Thankyou Kennedy! I knew I didn't invent the concept.

Half the confusion in this thread stems from the fact that the
question: "How large is this image?" has two answers.
1. a dimensionless pixel count (commensurate with no real world
measure)
2. a physical size, by combining the pixel count with the resolution
info.
Of course, for images that don't have an associated resolution, the
latter answer is not available.

I entered this thread to contradict jjs's assertion that file formats
do not encode resolution:

I never heard of DPI metric stored in an image

Seems he has some reading to do.

--Toby

Dimensionless units exist all over the place, just as dimensionless
constants do - pi, e, etc.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when

replying)

 

[gruhn]

It's a count. It's also dimensionless.

I see what you're saying. I see what I was saying. You're right. I was on
crack.

 

[Hecate]

A pixel count is dimensionless. The resolution value gives a pixel a
physical scale. There's really nothing to discuss here - this is all
taken as read.

Of course it isn't diemnsionless. It has a width and a height.

 

[Brian]

it were



Thankyou Kennedy! I knew I didn't invent the concept.

Half the confusion in this thread stems from the fact that the
question: "How large is this image?" has two answers.
1. a dimensionless pixel count (commensurate with no real world
measure)
2. a physical size, by combining the pixel count with the resolution
info.
Of course, for images that don't have an associated resolution, the
latter answer is not available.

I entered this thread to contradict jjs's assertion that file formats
do not encode resolution:


Seems he has some reading to do.

--Toby



replying)

Sorry to tell you this Toby, some image formats DO store resolution
(ppi) information. Corel Photopaint does. If I have an image that is
1800 x 1200 pixels, I can save it as a 6" x 4" at 300ppi, and a 2nd copy
as a 12" x 8" at 150dpi. I then import the 2 images into CorelDraw and
one image is 4 times the size of the other (twice the height, twice the
width) and both images are the same "pixel" size.

Brian.

 

[toby]

Sorry to tell you this Toby, some image formats DO store resolution
(ppi) information. ...

If you re-read the thread you'll see that that is EXACTLY what I was
trying to explain to someone else. These are the words I used:

"Many image formats -- from TIFF all the way down through BMP -- have
header fields specifically defined for resolution info".

 

[Brian]

If you re-read the thread you'll see that that is EXACTLY what I was
trying to explain to someone else. These are the words I used:

"Many image formats -- from TIFF all the way down through BMP -- have
header fields specifically defined for resolution info".

oops! I sometimes get confused as to who is saying what in here (being
new to NG's). My apologies Toby.

Brian

 

[Kennedy McEwen]

Of course it isn't diemnsionless. It has a width and a height.

Oh don't get us started on that again! Pixels are abstract objects
which do not have any intrinsic dimensions *until* they are physically
rendered. If they had a fixed width and height, how would you explain
the same image, with the same pixels, being printed at different sizes -
special relativity?

 

[MOP]

Of course it isn't diemnsionless. It has a width and a height.

Technically it is dimensionless when you write it you say 10 pixels (for
example) that does not have physical size, 1 pixel could be a mm or a mile
but would still be a pixel.
where as when it's DPI it dots/inch so has a physical size.
there are quite a few physical units that are dimension-less

 

[jjs]

Technically it is dimensionless when you write it you say 10 pixels (for
example) that does not have physical size, 1 pixel could be a mm or a mile
but would still be a pixel.
where as when it's DPI it dots/inch so has a physical size.
there are quite a few physical units that are dimension-less

It's rather like the US dollar; valueless until applied to a given market,
and then it's big or small.

 

[nobody]

Technically it is dimensionless when you write it you say 10 pixels

Depends. One school of thought says 10 pixels is different than 10
bananas, so while, for instance, it's valid to divide a quantity
expressed in pixels by another expressed as pixels per inch to get a
quantity expressed in inches, it doesn't quite work the same with
bananas.

 

[nobody]

Yes. Handwriting is nearer to a drawing,

No, it's _writing_. There was an age before print and computers and if
I am not mistaken people still use it every now and then to record and
relay information.

so I wouldn't count that unless
only the text itself needs to be recorded.

The question is can an OCR program recognize handwriting (reliably)?

There are plenty of markup
languages for equations.

The question is can an OCR program recognize mathematical equations
and symbols with ease?

Do you suggest nobody scibble those anymore but everybody carry
notebooks with them and use markup language instead?

In what way might a different language be a
problem?

Does your OCR recognize Arabic prose?

The point being, sometimes it's not feasible to OCR written text.
That's when you need to store or transmit such scanned documents as
images, so the orginal poster has a very valid need.

 

[jjs]

The question is can an OCR program recognize handwriting (reliably)?

No, and to head off those who say that the new tablet computers do a 'good
job', note that their method relies upon information not available in a
static bitmapped image, for example they sample characters as they are being
made - direction of the drawing of the character and turns - to find
statistical likelihood of the character being a,b,c etc.. And they still do
a bad job. Mathematical symbols are harder yet for advanced mathematicians -
if you knew the fights they have over accepted typography you would think
them nutz.