Why exactly does Paint.NET make bigger files than Irfanview?

On Wed, 15 Feb 2017 17:57:17 -0800, Savageduck wrote:

I am saying that whenever you save, resave, or save as a JPEG it is
subject to degradation of image quality regardless of the compression
settings made in the software

I can't argue what you're saying, but I will point out that Irfanview has
what they call a JPG "lossless" crop and rotation, so they must be special
cases in some way.

http://i.cubeupload.com/OymrtK.jpg

In article , Stijn De Jong
wrote:


I am saying that whenever you save, resave, or save as a JPEG it is
subject to degradation of image quality regardless of the compression
settings made in the software

I can't argue what you're saying, but I will point out that Irfanview has
what they call a JPG "lossless" crop and rotation, so they must be special
cases in some way.

they are.

"J. P. Gilliver (John)" wrote

| I've long suspected it to be as you say, but if that _is_ the case, what
| does the 100% (or 0) actually _mean_?

As far as I know it's an arbitrary scale. It's
not percentage, although that seems like a
reasonable assumption.

I don't really understand how it all works, except
that uniqueness between pixels is dropped out in
ways that the human eye tends not to notice, in
order to require less data to store the image pixels.
Or to put it another way, the image is degraded in
a very clever way to make it look as good as possible
despite the damage.

The 0-100 Quality scale does correspond to a
degree of compression, but the 0-100 range itself
seems to be just a convenience -- a handy numeric
range to convey the realitsic range of compression.
There's a chart he

https://en.wikipedia.org/wiki/JPEG

It correlates compression ratio with Quality factor.
But even that seems to be deceptive, because it's
lossy compression. So it's not really a factor of how
much compression there is but of how much data is
dropped out. If there's also a ZIP-style compression
operation involved I don't know what it is.

The article explains that the highest quality is encoded
at 8-9 bits per pixel. An actual displayed pixel is 24-bit,
so that seems to be about 2/3 of the data dropped out,
by reducing uniqueness. The 8-9 would be an average.
That can be clearly seen in the J.P.G. "gravestone" sample
images. At lower compression levels there's extreme
dithering. At each level there's less uniqueness in any
given area.

On 2017-02-16 02:48:51 +0000, Stijn De Jong said:

On Wed, 15 Feb 2017 17:57:17 -0800, Savageduck wrote:

I am saying that whenever you save, resave, or save as a JPEG it is
subject to degradation of image quality regardless of the compression
settings made in the software

I can't argue what you're saying, but I will point out that Irfanview has
what they call a JPG "lossless" crop and rotation, so they must be special
cases in some way.

http://i.cubeupload.com/OymrtK.jpg

That applies only to crop and/or rotation without additional
adjustment, editing or resizing, made to a rotated or cropped image
with the "JPG lossless" function applied.
This is one of those features which can work occassionally, but is
generally subject to the same lossy characteristics of all JPEG's.
Again, one would need to make a close examination of the
uncropped/unrotated original JPG against the "JPG Lossless
crop/rotation" result to determine the degree of success of the
Irfanview "JPG Lossless crop/rotation" claim. You might get away with
using that once or twice, but the failings of JPEG compression will
catch up with you.

--
Regards,

Savageduck

On Wed, 15 Feb 2017 21:19:14 -0500, Mayayana wrote:

I still don't really get it. It's the term itself.
Does it refer to the image as a 2-D space?
If I look in PSP I see resizing the canvas. It
seems to mean the same as creating a new
image and pasting the image onto that, which
is what I would do. So when Stijn refers to
"adding a canvas" does he mean that --
creating a larger background around the
original image?

Here is where I get the word "canvas" from.

When I need a colored (usually white) area on the side (usually bottom) of
an image for a caption, I can use either Irfanview or Paint.NET to create
that white space, both of which refer to the white space as a "canvas".

1. Irfanview Canvas:
Here is the Irfanview GUI to create that border white space:
http://i.cubeupload.com/CrZ318.jpg

2. Paint.NET Canvas:
Here is the Paint.NET GUI to create that border white space.
http://i.cubeupload.com/7NwPzt.jpg

Both Irfanview and Paint.NET call it the "canvas".
So I call it the canvas.

Do you call it something else?

"Stijn De Jong" wrote


| 2. This is the photo after Irfanview 80% batch resized it to 800x600 &
| Irfanview added a white canvas to the bottom:
| http://i.cubeupload.com/qfcHIm.jpg
| It is now 149 KB.

I see. I think it's deceptive to think of "adding
a canvas". IV has decompressed the JPG in order
to display it and then paints that on the larger
background that you specify. At that point it's a
bitmap that's been painted to a larger background
bitmap. There's no loss of data there.

Once you resave that it's going back to JPG so
you'll lose data. You then open that in PN and add
text, then save. PN is doing the same thing. It has
to decompress and render a bitmap in order to work
on the image, but then when you resave as JPG it's
compressing that bitmap, which is already an image
of a compressed image.

It's misleading to think of it as just an image that
you're moving here and there. As a bitmap that's
true. A bitmap is just that: a map of each pixel's
color values as numbers. When that's saved to JPG
the map is simplified so that it can take up less space.
But it's lossy. So when you then open the image again
you're opening a bitmap that's had data dumped. You
can't get that detail back. That happens every time
you work on it. Save a bitmap 20 times and it's still
the same bitmap. Save a JPG 20 times and you reduce
it to sludge.

If you could do the ops all at once you'd be better
off. Do all of the operations in PN, then save as TIF,
in case you want that image for something later.
(In this case TIF should be just a bitmap with lossless
compression. You can also save it as a BMP, but a
BMP can be shrunk to about 10% original size by
just applying ZIP compression, which doesn't change
the bitmap. So TIFs are the same thing but smaller.)

Then you can also save as JPG when you're done,
if necessary. That will give you the best possible quality
and you can then play with reducing JPG quality in order
to get a smaller file. Or if the size is not critical and
you'e putting it online, you can save to PNG. That
will shrink the file size without losing data. But it
won't get as small.

On Wed, 15 Feb 2017 21:10:47 -0500, nospam
wrote:

In article , Tony Cooper
wrote:

That works fine, especially since Paint.NET is one of the best
freebie
editors on Windows for texting correctly (there are a zillion ways
that other editors make texting hard)

paint.net does not text, nor does any other image editor. for that,
you need a dedicated text messaging app.

You could not figure out that a native speaker of a language other
the English meant "...for adding text (to an image)"?

that's called annotating, not texting.

texting has a well defined meaning which is not the same as
annotating.

Yes. It's obvious he made an error in the choice of his words. What
is also obvious is that there was no need to point it out.

there was.

I agree, there was a need to point it out. The need was to satisfy your
child like urge to comment.
As you have so often said, "it was "picking on words."

He doesn't like it when his are picked on, but jumps in to pick on the
word's of others when the wrong one is chosen.

says the person who jumps on the words of others, particularly me,
going so far to intentionally lie and twist what i say solely to argue
and criticize it, just as you are doing here.

when others make 'an obvious error', you give them a free pass and then
criticize me for doing what you normally do.

Sure. I give non-native speakers of English a free pass*, You should
know better. Try responding to the OP in Dutch.

and in fact, i do know better.

as usual, you are butting in just to attack.

the person in question is a well known troll who lives in the santa
cruz mountains and is a native english speaker. he regularly spoofs his
address, however, his style is unmistakable and easily outed.

both savageduck and myself know him quite well from other newsgroups.

Well, then, I apologize. I wasn't aware of that.

Why, if you knew this, did you even bother to reply?


*Unless they brag about how good their English is.

What lie? What was twisted?

everything.

No, no lie, nothing twisted, just a lack of knowledge about the
backstory.

--
Tony Cooper - Orlando, Florida

"Stijn De Jong" wrote
|
| I am saying that whenever you save, resave, or save as a JPEG it is
| subject to degradation of image quality regardless of the compression
| settings made in the software
|
| I can't argue what you're saying, but I will point out that Irfanview has
| what they call a JPG "lossless" crop and rotation, so they must be special
| cases in some way.

That's an interesting one. IV must be altering
the data storage to do something like turn the
orientation 90 degrees without changing the image
data. But that's a special case. In general, any
JPG save will lose data.

On Wed, 15 Feb 2017 17:56:19 -0500, Mayayana wrote:

Whaddaya mean, resized?

In the original post, I "resized" the files.
In the baseline test, I did NOT resize the files.

The reason I didn't resize the files in the baseline test is that I had
*expected* the size from Paint.NET alone to be larger in all cases (it
wasn't - but that was my test).

In the original post, I do what I do all the time:
1. I start with a set of original JPEG files from the phone.
2. In one step, I use Irfanview to batch lossless rotate & reset EXIF
orientation
3. In the second batch step, I use Irfanview to batch resize them to
800x600 pixels and to add a 50-pixel white canvas to the bottom.
4. Since all other known photo-editing programs, including Irfanview, stink
for adding text, circling areas, and for adding curved and dotted arrows, I
use Paint.NET for annotating the photographs (often they are screenshots).

I find when I am done with Step 4. the file sizes are as much as twice as
large as they were at the end of Step 3!

So I run a Step 5, which is simply to batch resize the files in Irfanview
again.

I assumed these were
all images of the same pixel dimensions, but resaved
and with added text. (I'm not sure what you mean by
adding a canvas. A monotone rectandgle for text?)

This is Usenet so it's hard for me to describe things to you that I do all
day, every day - so I'll show you what I mean by adding a canvas simply by
referrring to these screenshots which I just did for a thread in the mobile
phone group (where we are looking up how to test the signal strength of the
local cellular tower):

01 Network Cell Info Lite, version 3.30:
http://i.cubeupload.com/HoKTav.jpg
http://wilysis.com/networkcellinfo
https://play.google.com/store/apps/d...s.cellinfolite

02 Network Signal Info, version 3.63.01:
http://i.cubeupload.com/2zK8Ys.jpg
https://play.google.com/store/apps/d...android.telnet

03 GSM Signal Monitoring, version 4.02:
http://i.cubeupload.com/V9O0Gg.jpg
https://play.google.com/store/apps/d...almonitorin g

04 Netmonitor, version 1.2.15:
http://i.cubeupload.com/TfDJaS.jpg
https://play.google.com/store/apps/d...ene.netmonitor

05 CellID Info:, version 1.2.2:
http://i.cubeupload.com/X3gsfb.jpg
https://play.google.com/store/apps/d...ere.cellidinfo

06 RF Toolbox (Cell Monitor), version 2.26:
http://i.cubeupload.com/y2YfEV.jpg
https://play.google.com/store/apps/d...apps.rftoolbox

07 WiGle WiFi Wardriving (which also reports cellular towers):
http://i.cubeupload.com/ZPva3O.jpg
https://play.google.com/store/apps/d...e.wigleandroid

08. OpenSignal, version 5.10:
http://i.cubeupload.com/BwfSFa.jpg
https://opensignal.com/app/
https://play.google.com/store/apps/d...se3.opensignal
etc.

I didn't put any arrows on those pictures, but everything else I did, which
was batch resize and then batch add a canvas, and then annotate
individually using Paint.NET, and then batch resize again in Irfanview (to
get the file size back to reasonable after saving with Paint.NET).

01 Network Cell Info Lite, version 3.30:
http://i.cubeupload.com/HoKTav.jpg
http://wilysis.com/networkcellinfo
https://play.google.com/store/apps/d...s.cellinfolite

02 Network Signal Info, version 3.63.01:
http://i.cubeupload.com/2zK8Ys.jpg
https://play.google.com/store/apps/d...android.telnet

03 GSM Signal Monitoring, version 4.02:
http://i.cubeupload.com/V9O0Gg.jpg
https://play.google.com/store/apps/d...almonitorin g

04 Netmonitor, version 1.2.15:
http://i.cubeupload.com/TfDJaS.jpg
https://play.google.com/store/apps/d...ene.netmonitor

05 CellID Info:, version 1.2.2:
http://i.cubeupload.com/X3gsfb.jpg
https://play.google.com/store/apps/d...ere.cellidinfo

06 RF Toolbox (Cell Monitor), version 2.26:
http://i.cubeupload.com/y2YfEV.jpg
https://play.google.com/store/apps/d...apps.rftoolbox

07 WiGle WiFi Wardriving (which also reports cellular towers):
http://i.cubeupload.com/ZPva3O.jpg
https://play.google.com/store/apps/d...e.wigleandroid

08. OpenSignal, version 5.10:
http://i.cubeupload.com/BwfSFa.jpg
https://opensignal.com/app/
https://play.google.com/store/apps/d...se3.opensignal
etc.

In article , Tony Cooper
wrote:


Sure. I give non-native speakers of English a free pass*, You should
know better. Try responding to the OP in Dutch.

and in fact, i do know better.

as usual, you are butting in just to attack.

the person in question is a well known troll who lives in the santa
cruz mountains and is a native english speaker. he regularly spoofs his
address, however, his style is unmistakable and easily outed.

both savageduck and myself know him quite well from other newsgroups.

Well, then, I apologize. I wasn't aware of that.

ok

Why, if you knew this, did you even bother to reply?

many reasons, one being that he used incorrect terminology.

In article , Mayayana
wrote:

| I am saying that whenever you save, resave, or save as a JPEG it is
| subject to degradation of image quality regardless of the compression
| settings made in the software
|
| I can't argue what you're saying, but I will point out that Irfanview has
| what they call a JPG "lossless" crop and rotation, so they must be special
| cases in some way.

That's an interesting one. IV must be altering
the data storage to do something like turn the
orientation 90 degrees without changing the image
data. But that's a special case. In general, any
JPG save will lose data.

it's a simple transform that does not require decompressing and
recompressing, thus lossless.

J. P. Gilliver (John) wrote:
In message 20170215163740260-savageduck1@REMOVESPAMmecom, Savageduck
writes:
On 2017-02-16 00:03:01 +0000, "J. P. Gilliver (John)"
said:

In message , Mayayana
writes:
[]
try saving to JPG and see what you get. If all of
your saves are to JPG then you're not dealing with
the same image each time, even if you set JPG
compression at "100". (Or 0, depending on the
software.) JPG always loses data with each save.

I've long suspected it to be as you say, but if that _is_ the case,
what does the 100% (or 0) actually _mean_?

It means that the compression algorithm of whatever software is in use
will not apply additional compression beyond the lossyness to be found
with each save/resave. Unfortunately, JPEG being what it is, is always
going to be subject to a degree of data and IQ loss on each
save/resave. Using a 100% or 0% (depending on software) compression
setting will still result in data loss, and will more than likely have
very visible JPEG compression artifacts present.

Do that enough times and add in resizing, the degradation in the JPEG
will be unacceptable to all but the least critical eye, even if the
file size might have grown.

So are you saying there are two different _sorts_ of data compression
applied when you save a JPEG?

The tombstone pictures about 75% down this web page,
show the tradeoff of compression ratio to image quality.
JPEG uses DCT (discrete cosine transform), to do a
frequency domain analysis of an image. The highest
frequencies can be discarded, on "non-cartoon images",
to great effect, so you can get a decent compression
ratio, while fooling the human eye into thinking
there is little degradation.

https://en.wikipedia.org/wiki/Jpeg

"More modern designs such as JPEG 2000 and JPEG XR
exhibit a more graceful degradation of quality as
the bit usage decreases – by using transforms with
a larger spatial extent for the lower frequency
coefficients and by using overlapping transform
basis functions."

One thing JPEG doesn't like, is "sharp edges".

Things similar to cartoons, work best with GIF
(which uses a lossless compressor).

Any time you have "content" in hand, you have to
select the best format for the content type. And JPEG
isn't always my favorite (ringing on sharp edges).

Paul

"Stijn De Jong" wrote

| Here is where I get the word "canvas" from.
|
| When I need a colored (usually white) area on the side (usually bottom) of
| an image for a caption, I can use either Irfanview or Paint.NET to create
| that white space, both of which refer to the white space as a "canvas".

| Do you call it something else?

Canvas seems fine. I just never noticed the term
before. If I'm adding a white stripe I would paste
the image onto a larger white image and merge the
two. If I need a white stripe in the existing image
I'd paint it with a shapes tool. I guess I've never
conceptually thought of the abstraction of a canvas
that holds the image. I'm always thinking in terms
of a bitmap because in actual practice that's what it
always is. "Adding canvas" would be accomplished by
painting a bitmap onto a second larger bitmap, just
because that's how Windows graphics works.

"Stijn De Jong" wrote

| Whaddaya mean, resized?
|
| In the original post, I "resized" the files.
| In the baseline test, I did NOT resize the files.
|
| The reason I didn't resize the files in the baseline test is that I had
| *expected* the size from Paint.NET alone to be larger in all cases (it
| wasn't - but that was my test).

I think we're "crossing in the mail" here. I figured
out what you were talking about. Though I don't
know why PN might be resaving significantly bigger
when all you've done was to add text.

"nospam" wrote

| That's an interesting one. IV must be altering
| the data storage to do something like turn the
| orientation 90 degrees without changing the image
| data. But that's a special case. In general, any
| JPG save will lose data.
|
| it's a simple transform that does not require decompressing and
| recompressing, thus lossless.

Not simple at all. To be lossless it has to
unpack and rearrange the encoded image
data without going through a step of unpacking
to bitmap and re-encoding. But if it's simple
for you I'd be very interested to see your code
to accomplish it. I'm guessing it's quite a trick.