Irfanview color depth

[Peter]

In Irfanview, you can Decrease Color Depth to any level but how can you
either increase it or figure out what the current color depth is set to?

[Newyana2]

"Peter" <***@nospam.net> wrote

| In Irfanview, you can Decrease Color Depth to any level but how can you
| either increase it or figure out what the current color depth is set to?

If you haven't changed anything then you'll mostly know
by file type. JPGs will be 24-bit, GIFs are 8-bit, PNGs
32-bit. 24-bit is actually the max in standard usage.
The extra 8 bits in PNGs are for transparency values.
BMPs can be 2-bit to 24-bit, but most people don't
see BMPs much these days.

Do you know what color depth is? A monitor these days
displays 24-bit color, which means 256 different hues of
red, green and blue. 0-0-0 is black. 255-255-255 is white.
(Unless you're on a Mac, in which case I think it's 18-bit
color. Which means it can only display 64 color gradients for
R, G and B -- it's missing 192 hues out of 256, so it dithers
pixels to the nearest color.)

Raster image file formats store those values as numbers
representing a pixel grid. They're all bitmap when they
display, but each format stores the data differently.

Why would you want to change color depth? If you want
to do something like save a JPG as GIF then you'll lose a lot
of the color data. If you do the reverse you won't get more
colors. The original GIF color table entries are all that you'll
see unless you then edit the image in 24-bit.

[Peter]

Post by Newyana2
Do you know what color depth is?

Not really. That's why I had asked the question to figure out how modifying
the image in various ways to increase entropy affected the color depth.

Post by Newyana2
Why would you want to change color depth?

To increase entropy.

Post by Newyana2
If you want
to do something like save a JPG as GIF then you'll lose a lot
of the color data.

Does saving JPG to GIF remove unique camera sensor imperfections?

[Newyana2]

"Peter" <***@nospam.net> wrote

| > Why would you want to change color depth?
|
| To increase entropy.
|
I looked that up. I don't find any reference to entropy
in graphics. So I'm not sure what you mean.

| > If you want
| > to do something like save a JPG as GIF then you'll lose a lot
| > of the color data.
|
| Does saving JPG to GIF remove unique camera sensor imperfections?

It will be "dithered" to nearest colors, based on one
of a number of dithering approaches. For example, if
you have a gradient of greens containing 372 hues, it
might get converted to a field of green and white dots.
A GIF can only use 256 colors, so most JPGs would be
severely degraded when saved as GIF, because a JPG
can use 16 million colors.

It helps to understand the basic system of raster images.
In ALL cases they represent pixel grids with numeric RGB
values. That is, all formats store data to light pixels on a
screen with varying intensities of red, green and blue. It's
always a grid, always rectangular. Any image that's not a
rectangle (like an icon or PNG) is still a rectangular bitmap,
but the image format is stroring transparency values to be
applied when the image is rendered onscreen. (That's why
displays are often "32-bit". 3 bytes for RGB values and one
byte to indicate transparency.)

The grid of stored pixel values, indicating color intensity of
RGB, is arranged in rows, usually starting at top left. That's
a bitmap. All raster images are bitmaps but can be stored
in different file formats. JPG is arguably a very poor format
because it loses color data, but it's popular because it makes
the smallest files and there are no royalties on the format.
So it's ideal for online. (It's used in cameras for that reason.
For people sending birthday party pictures in email, quality
is not a big factor.)

Once you open an image in a graphic editor you're dealing
with the bitmap, so whatever you alter from there will affect
the image saved as a different file. In the case of JPG, it
compresses the image by eliminating contiguous colors in
imperceptible ways. That's why a bad quality JPG looks like
an image comprised of blocks. The reduction of colors allows
for the data to be stored more compactly, but loses detail.
That data is lost for good.

So if you have a JPG saved at, say, 92 compression (it's
1 to 100. Top quality can be either the high or low number,
depending on the software), then if you open that in an editor
and resave it at 87 compression, there should be no noticeable
difference, but some byte values will be changed. (I like Paint
Shop Pro 5 because it chops off the EXIF data altogether. I
find it creepy to have buried data in a file. It's a privacy
problem.)

So if it were me I'd try resaving the JPG at different
compression, convert both of those to BMP, then compare in
a hex editor to see what you have. If you convert to GIF
you'll ruin the image because it has to dither to a max of 256
colors, while the JPG could have 100,000 colors. So forget GIF.

If you open a JPG in a hex editor it won't be very informative.
It's like looking at a ZIP file. You only see a bloated header and
the compressed state of the data.

If you open and resave as a BMP then you
have the direct data. The first 54 bytes of the BMP file will
contain values indicating color depth, width/height, etc. The
rest is simply the straight grid values. So for a typical 24-bit
BMP image, the first 3 bytes will be the BGR values in big-endian
order for the top left pixel. Example: Bright sky blue is zero red,
half green intensity, and full blue intensity. As a long integer
value that's 16744448. As bytes it's 255-128-0 or 0-128-255.
That can also be written as hex: FF 80 00 If you save a BMP
file which is only that color then you'll see 54 bytes of file header
followed by a repeating pattern of FF 80 00.

All raster images work that way. All raster images are bitmaps
in different packaging. A JPG is also a bitmap, but when you
increase compression you'll reduce colors. So if you have, say, a
photo of sky with pixels like 255 80 00 243 75 22 241 83 02 those
three pixels might get dithered to 3 pixel values of 243 75 22. Your
eye won't see the difference, but the 3 pixels' values can be more
easily compressed.

So you could try that. Check compression level, open the file,
resave at different compression, open both files and resave as
BMPs. Open both BMPs in a hex editor and see how they compare.

I can't tell you anything about camera sensors. I don't know about
that. But however they work, it still has to boil down to 24-bit
RGB if you have a JPG. So any tracks left by the camera would
have to be in patterns of pixel values.

I hope that makes sense. It sounds complicated, but it's actually
very simple once you get how it orks. All raster images are grids
of pixel RGB values as numbers. It all comes down to numbers,
just as any file does.

[Peter]

Post by Newyana2
| To increase entropy.
|
I looked that up. I don't find any reference to entropy
in graphics. So I'm not sure what you mean.

Thanks for spending the effort to understand the reason for caring about
entropy in terms of digital forensics of images posted to online sources.

Entropy is a common fundamental technical term for levels of "disorder."
Images: https://duckduckgo.com/?q=camera+fingerprinting+%2Bentropy
Smartphones: https://duckduckgo.com/?q=smartphone+sensor+fingerprinting+%2Bentropy
Browsers: https://duckduckgo.com/?q=browser+fingerprinting+%2Bentropy
Digits: https://duckduckgo.com/?q=fbi+fingerprinting+%2Bentropy

I'm using the term the way they use it to uniquely identify every image
posted to the Internet that came from any particular unique camera sensor.

For example, this article on "Smartphone Camera Identification" discusses
entropy 25 times and camera 114 times (where it's almost a 1:4 ratio) and
where they said "In this work, we follow an identification methodology for
smartphone camera sensors... Our analysis showed that the blue channel
provided the best separation..."
https://www.mdpi.com/1099-4300/24/8/1158/html
https://mdpi-res.com/d_attachment/entropy/entropy-24-01158/article_deploy/entropy-24-01158-v3.pdf

And this paper titled "Mobile Device Identification via Sensor
Fingerprinting" uses the words entropy:camera in a 3:2 ratio, where they
conclude "We show that the entropy from sensor fingerprinting is sufficient
to uniquely identify a device."
https://www.arxiv-vanity.com/papers/1408.1416/

The problem of entropy in terms of posting images to social media is
described in this paper on "Robustness of digital camera identification"
https://link.springer.com/article/10.1007/s11042-021-11129-y
Where they start off with "One of the problem in digital forensics is the
issue of identification of digital cameras based on images. This aspect has
been attractive in recent years due to popularity of social media platforms
like Facebook, Twitter etc., where lots of photographs are shared."

Given this conclusion from the following paper on image fingerprinting
"No Two Digital Cameras Are the Same: Fingerprinting Via Sensor Noise"
https://33bits.wordpress.com/2011/09/19/digital-camera-fingerprinting/
"Camera fingerprinting can be used on the one hand for detecting forgeries
(e.g., photoshopped images), and to aid criminal investigations by
determining who (or rather, which camera) might have taken a picture. On
the other hand, it could potentially also be used for unmasking individuals
who wish to disseminate photos anonymously online."

Let's make up a scenario where it might matter (but please don't shoot the
example but try to understand the problem the example is illustrating).

a. You are brought up Christian & you post to your local church site
b. You are an employee & you post images to your employer web site
c. You have political aspirations & you post images to your party web site
d. You are LGBTQ+ & you post images to your favorite LGBTQ+ web site

Do you want all those online photos to uniquely identify your camera?

Post by Newyana2
|> If you want
|> to do something like save a JPG as GIF then you'll lose a lot
|> of the color data.
|
| Does saving JPG to GIF remove unique camera sensor imperfections?
It will be "dithered" to nearest colors, based on one
of a number of dithering approaches. For example, if
you have a gradient of greens containing 372 hues, it
might get converted to a field of green and white dots.
A GIF can only use 256 colors, so most JPGs would be
severely degraded when saved as GIF, because a JPG
can use 16 million colors.

According to one of the papers above, the "blue" realm is the easiest to
fingerprint (although some papers indicated it was the "green").

This JPG-to-GIF dithering might therefore help in increasing entropy.

Post by Newyana2
It helps to understand the basic system of raster images.

What I do not understand which is important is how the camera's sensor
imperfections show up in the camera's resulting outputted raster images.

Post by Newyana2
In ALL cases they represent pixel grids with numeric RGB
values. That is, all formats store data to light pixels on a
screen with varying intensities of red, green and blue. It's
always a grid, always rectangular. Any image that's not a
rectangle (like an icon or PNG) is still a rectangular bitmap,
but the image format is storing transparency values to be
applied when the image is rendered onscreen. (That's why
displays are often "32-bit". 3 bytes for RGB values and one
byte to indicate transparency.)

It would seem that the fewest bits used (which show the image with just
enough clarity to be useful) would be the best to increase entropy.

Post by Newyana2
The grid of stored pixel values, indicating color intensity of
RGB, is arranged in rows, usually starting at top left. That's
a bitmap. All raster images are bitmaps but can be stored
in different file formats. JPG is arguably a very poor format
because it loses color data, but it's popular because it makes
the smallest files and there are no royalties on the format.
So it's ideal for online. (It's used in cameras for that reason.
For people sending birthday party pictures in email, quality
is not a big factor.)

For the reasons you stated, most images online are JPG so that's what I'm
trying to increase the entropy of. If an automatic JPG->GIF->JPG operation
for all uploaded files increases that entropy, then that's probably a good
technique that I can use to hinder fingerprinting by increasing entropy.

Post by Newyana2
Once you open an image in a graphic editor you're dealing
with the bitmap, so whatever you alter from there will affect
the image saved as a different file. In the case of JPG, it
compresses the image by eliminating contiguous colors in
imperceptible ways. That's why a bad quality JPG looks like
an image comprised of blocks. The reduction of colors allows
for the data to be stored more compactly, but loses detail.
That data is lost for good.

The ultimate web site might perform that image alteration also.

I don't know what they do with the images though so I don't have control
over whether they increase the entropy further or leave it alone.

Post by Newyana2
So if you have a JPG saved at, say, 92 compression (it's
1 to 100. Top quality can be either the high or low number,
depending on the software), then if you open that in an editor
and resave it at 87 compression, there should be no noticeable
difference, but some byte values will be changed. (I like Paint
Shop Pro 5 because it chops off the EXIF data altogether. I
find it creepy to have buried data in a file. It's a privacy
problem.)

It would be nice to know how much JPEG compression alone increases (or
decreases) entropy. I would assume it increases entropy.

But I have no idea if it's a lot or only a little.
That they uniquely identify cameras from images hints at little.

Post by Newyana2
So if it were me I'd try resaving the JPG at different
compression, convert both of those to BMP, then compare in
a hex editor to see what you have. If you convert to GIF
you'll ruin the image because it has to dither to a max of 256
colors, while the JPG could have 100,000 colors. So forget GIF.

I'll try the JPG->GIF->JPG method to see if it "ruins" the image.

Post by Newyana2
If you open a JPG in a hex editor it won't be very informative.
It's like looking at a ZIP file. You only see a bloated header and
the compressed state of the data.

Yes but that data is very informative when it uniquely identifies your
camera out of pictures scattered across web sites on the Internet.

Post by Newyana2
If you open and resave as a BMP then you
have the direct data. The first 54 bytes of the BMP file will
contain values indicating color depth, width/height, etc. The
rest is simply the straight grid values. So for a typical 24-bit
BMP image, the first 3 bytes will be the BGR values in big-endian
order for the top left pixel. Example: Bright sky blue is zero red,
half green intensity, and full blue intensity. As a long integer
value that's 16744448. As bytes it's 255-128-0 or 0-128-255.
That can also be written as hex: FF 80 00 If you save a BMP
file which is only that color then you'll see 54 bytes of file header
followed by a repeating pattern of FF 80 00.

The less repeatable (more random) each image's bitmapped digital result is
on the online storage medium, is, the better for increasing entropy.

Post by Newyana2
All raster images work that way. All raster images are bitmaps
in different packaging. A JPG is also a bitmap, but when you
increase compression you'll reduce colors. So if you have, say, a
photo of sky with pixels like 255 80 00 243 75 22 241 83 02 those
three pixels might get dithered to 3 pixel values of 243 75 22. Your
eye won't see the difference, but the 3 pixels' values can be more
easily compressed.

I like that increasing compression reduces colors. What you want to do, I
would think, is paper over the camera sensor imperfections in the output.

Post by Newyana2
So you could try that. Check compression level, open the file,
resave at different compression, open both files and resave as
BMPs. Open both BMPs in a hex editor and see how they compare.

What I'll test is JPG->GIF->JPG and JPG->BMP->JPG to see which gives the
best results for an online upload - but which do you think introduces the
most entropy?

Post by Newyana2
I can't tell you anything about camera sensors. I don't know about
that. But however they work, it still has to boil down to 24-bit
RGB if you have a JPG. So any tracks left by the camera would
have to be in patterns of pixel values.

The articles I pointed to in the beginning of this response shows that what
digital forensics target are the camera sensor's unique imperfections.

Post by Newyana2
I hope that makes sense. It sounds complicated, but it's actually
very simple once you get how it orks. All raster images are grids
of pixel RGB values as numbers. It all comes down to numbers,
just as any file does.

Thank you for all your helpful information. The goal is to introduce "just
enough" entropy so that all your images aren't uniquely traced to you.

[Newyana2]

"Peter" <***@nospam.net> wrote

| Thank you for all your helpful information. The goal is to introduce "just
| enough" entropy so that all your images aren't uniquely traced to you.

Interesting issue. I'd never heard of it. I wasn't up to fully
reading all those articles, but I have the basic idea. Whether
it will ever be a way to ID the picture taker from an image
is questionable. It seems a bit like doing a DNA test on
someone wearing a name tag. The site already knows who's
uploading in most cases, and cameras are leaking data that allows
for ID. But I can see how this could someday be an issue.

As for introducing entropy, I don't know. There's no sense
having images that are ruined, so you don't want to change
too much. And this doesn't look like entropy to me. Rather, they're
loking for identifiable patterns of distortion. What would that be?
Maybe the sensor never reports certain hue values? I don't know.
I think you'll just have to test, after coming up with some way
to gauge how unique the ID traces are. You don't know if your
method is helping unless you know what to look for.

JPG -> GIF -> JPG will ruin nearly all images.
Any JPG resaving will change the image because it's
a lossy format. But other formats are not lossy. So it's
in the JPG resaving that you'll get the most change.

The BMP saving would only be for inspecting byte changes
for pixel values. There's no other advantage to BMP. The
image displayed is already a BMP, anyway. It's the display
of what's called a DIB -- device independent bitmap. That
is, just the pixel bytes.

So resave your JPG, stripping the header and reducing
quality slightly. Then see what you have by saving that
as a BMP. But I have no idea how you'll assess the uniqueness.
If you had some formula for that you could probably
automate it by processing the byte value patterns. But
that means getting some source code for the software that
will supposedly do the job. In other words, if yopu build up
an ID for your camera from multiple images then you can
test your alteration against that, but without having that,
I don't know how you'll identify exactly what bytes in the
image are giving you away, and why.

[Peter]

Post by Newyana2
Interesting issue. I'd never heard of it.

Remember a decade ago you could post a picture & nobody would be able to
find that you posted it in another place?

Now they can. And it's just going to get worse with compute power.

Post by Newyana2
I wasn't up to fully
reading all those articles, but I have the basic idea. Whether
it will ever be a way to ID the picture taker from an image
is questionable. It seems a bit like doing a DNA test on
someone wearing a name tag. The site already knows who's
uploading in most cases, and cameras are leaking data that allows
for ID. But I can see how this could someday be an issue.

The more powerful computers get and the cheaper mass storage is, the more
it will be feasible for the powers that be to uniquely identify all the
images ever posted to a variety of diverse locations to each camera.

Post by Newyana2
As for introducing entropy, I don't know. There's no sense
having images that are ruined, so you don't want to change
too much. And this doesn't look like entropy to me. Rather, they're
loking for identifiable patterns of distortion. What would that be?

I agree they seem to be looking for identifiable pixel flaws in the sensor
output which show up in the same spots on every image that you upload.

Post by Newyana2
Maybe the sensor never reports certain hue values? I don't know.
I think you'll just have to test, after coming up with some way
to gauge how unique the ID traces are. You don't know if your
method is helping unless you know what to look for.

You can move the spots around by cropping & tilting but some spots will
remain unless you know exactly where all are so as to crop them all out.

Post by Newyana2
JPG -> GIF -> JPG will ruin nearly all images.

I just tried it on a bunch of images and they're not ruined for the purpose
of uploading them to a typical web site (this isn't professional stuff).

Luckily Irfanview can batch convert so I just ran the "b" command a couple
of times to convert from JPEG to GIF and back to JPEG. The quality was ok.

Post by Newyana2
Any JPG resaving will change the image because it's
a lossy format. But other formats are not lossy. So it's
in the JPG resaving that you'll get the most change.

Thanks for that advice. In the JPG->GIF->JPG, the original photo had 219908
unique colors, while the final uploaded photo had 80855 unique colors.

Original (as reported by Irfanview "i" command):
Original colors = 16.7 Million (24 BitsPerPixel)
Current colors = 16.7 Million (24 BitsPerPixel)
Number of unique colors = 232666

GIF (as reported by Irfanview "i" command):
Original colors = 256 (8 BitsPerPixel)
Current colors = 256 (8 BitsPerPixel))
Number of unique colors = 256

Uploaded (as reported by Irfanview "i" command):
Original colors = 16.7 Million (24 BitsPerPixel)
Current colors = 16.7 Million (24 BitsPerPixel)
Number of unique colors = 80801

It dropped the number of unique colors by an order of magnitude.

Interestingly, running Irfanview autoadjust colors didn't change much.
Original colors = 16.7 Million (24 BitsPerPixel)
Current colors = 16.7 Million (24 BitsPerPixel)
Number of unique colors = 80375

But oh what a difference in unique colors Irfanview sharpen did!
Original colors = 16.7 Million (24 BitsPerPixel)
Current colors = 16.7 Million (24 BitsPerPixel)
Number of unique colors = 169885

Why would a sharpen add so many unique colors to the image?

Running an Irfanview "Effects -> Blur" didn't change all that much.
Original colors = 16.7 Million (24 BitsPerPixel)
Current colors = 16.7 Million (24 BitsPerPixel)
Number of unique colors = 146437

But of course, I have no way of knowing if this papers over unique camera
sensor flaws that were in the original image that moved forward throughout.

Post by Newyana2
The BMP saving would only be for inspecting byte changes
for pixel values. There's no other advantage to BMP. The
image displayed is already a BMP, anyway. It's the display
of what's called a DIB -- device independent bitmap. That
is, just the pixel bytes.

Since the DISPLAY is a BMP, would you think it useful to paper over unique
camera sensor flaws by snapping a screenshot and replacing it with that?

Post by Newyana2
So resave your JPG, stripping the header and reducing
quality slightly. Then see what you have by saving that
as a BMP. But I have no idea how you'll assess the uniqueness.

I think we need to fundamentally do a few things, but I'm not sure.

We need to paper over the unique flaws with blurring somehow.
And maybe we need to paper over them with color changes somehow.
And maybe we can move them around by cropping & tilting the images.

Post by Newyana2
If you had some formula for that you could probably
automate it by processing the byte value patterns. But
that means getting some source code for the software that
will supposedly do the job.

They generally test it with a black photo but I think they only do that so
as to have consistent input for their algorithms to assess sensor flaws.

Post by Newyana2
In other words, if yopu build up
an ID for your camera from multiple images then you can
test your alteration against that, but without having that,
I don't know how you'll identify exactly what bytes in the
image are giving you away, and why.

Yup. I know that they're looking for unique flaws, but I don't know how to
run the Irfanview "i" command to find the entropy of any given image file.

[Newyana2]

"Peter" <***@nospam.net> wrote

| > JPG -> GIF -> JPG will ruin nearly all images.
|
| I just tried it on a bunch of images and they're not ruined for the
purpose
| of uploading them to a typical web site (this isn't professional stuff).
|

I tried a few myself. Surprisingly they look fine!
One is a photo of a woodcock sitting amongst
greenery. Lots of hues. Yet the GIF looks OK. And
as you noted, the color count increases when it's
converted back. Somehow the JPG conversion brings
back some sharpness.

| Why would a sharpen add so many unique colors to the image?
|
Sharpen highlights the difference along edges. If you
take a simple image -- a single color background with
a different color rectangle in the middle, say, a sharpen
will add 1 or more lies of new colors where the 2 colors
meet. In a photo you're doing that with a slightly different
hue for each pixel comparison.

| But of course, I have no way of knowing if this papers over unique camera
| sensor flaws that were in the original image that moved forward
throughout.
|

No. You should have virtually all new pixels, but I don't
know how their method works.

| Since the DISPLAY is a BMP, would you think it useful to paper over unique
| camera sensor flaws by snapping a screenshot and replacing it with that?
| >

A screenshot of the desktop? That should give you
just the same bitmap. A screenshot will just send you the
byte values being sent to the graphic hardware.

| I think we need to fundamentally do a few things, but I'm not sure.
|

Maybe put it in a locked box, sseal that with wax, then
bury it under your garage floor. :)

| We need to paper over the unique flaws with blurring somehow.

Are they flaws? Could they be something like particular
hues that can never show up? I don't know. without knowing
the method of inspection you're in the dark. But I'm guessing
your method will have changed the precise hue values of nearly
every pixel. So... pretty much a new image that just happens
to look the same to your eye.

[Peter]

Post by Newyana2
I tried a few myself.

Thank you for trying that out as whatever we come up with to increase
entropy and paper over unique camera sensor flaws must be easy to do.

It should probably be all done inside of Irfanview, if that's possible.

One way we can probably check our entropy is to pick a photo from the
Internet and then run our commands on that, & then upload it somewhere.

A few weeks later it should make it into the reverse-image-search engines.

Post by Newyana2
Surprisingly they look fine!

I agree. For an upload, they work well (bearing in mind I also resize).

I like that the Irfanview JPG->GIF adds disorder (in the form of collapsing
the unique colors) and then the Irfanview batch reconversion to JPG adds
them back - so the powers that be might not notice it had been done.

Post by Newyana2
One is a photo of a woodcock sitting amongst greenery. Lots of hues.

Your idea of JPG->GIF->JPG was wonderful, which I had never thought of
before you mentioned it in a couple of posts ago where the conversion needs
to be quick & easy and all done inside of one program, Irfanview.

To that end, I was re-testing your JPG->GIF->JPG ideas this morning (with
more consistent settings, like no compression, cropping, tilting &
sharpening) when I realized there may be two easy ways in Irfanview to
accomplish that intermediate GIF conversion of 200K colors to 256 colors.

One way is the way we've been discussing, which is a batch convert of JPG
to GIF to JPG with a resharpen added to greatly increase the number of
colors to a naturally unsuspicious level for typical uploaded JPG images.

But the other way "might" be to just keep it in JPG but reduce the colors
Irfanview -> Image -> Reduce color depth -> 256 colors (8 BPP)
(which is what I had been doing that kicked off this original question).

[1] Original photo
Compression: JPEG, quality: 96, subsampling ON (2x2)
Original size: 4000 x 3000 Pixels (12.00 MPixels) (4:3)
Current size: 4000 x 3000 Pixels (12.00 MPixels) (4:3)
Print size (from DPI): 141.1 x 105.8 cm; 55.56 x 41.67 inches
Original colors: 16,7 Million (24 BitsPerPixel)
Current colors: 16,7 Million (24 BitsPerPixel)
Number of unique colors: 219908
Disk size: 4.48 MB (4,693,172 Bytes)

[2] JPG->GIF (using default Irfanview "b" settings)
Compression: GIF - LZW
Original size: 4000 x 3000 Pixels (12.00 MPixels) (4:3)
Current size: 4000 x 3000 Pixels (12.00 MPixels) (4:3)
Print size (from DPI): 141.1 x 105.8 cm; 55.56 x 41.67 inches
Original colors: 256 (8 BitsPerPixel)
Current colors: 256 (8 BitsPerPixel)
Number of unique colors: 256
Disk size: 7.06 MB (7,401,292 Bytes)

GIF->JPG (using default Irfanview "b" settings)
Compression: JPEG, quality: 100, subsampling ON (2x2)
Original size: 4000 x 3000 Pixels (12.00 MPixels) (4:3)
Current size: 4000 x 3000 Pixels (12.00 MPixels) (4:3)
Print size (from DPI): 141.1 x 105.8 cm; 55.56 x 41.67 inches
Original colors: 16,7 Million (24 BitsPerPixel)
Current colors: 16,7 Million (24 BitsPerPixel)
Number of unique colors: 98371
Disk size: 9.17 MB (9,615,126 Bytes)

[3] Reduce colors using Image -> Reduce color depth -> 256 colors (8 BPP)
Compression: JPEG, quality: 96, subsampling ON (2x2)
Original size: 4000 x 3000 Pixels (12.00 MPixels) (4:3)
Current size: 4000 x 3000 Pixels (12.00 MPixels) (4:3)
Print size (from DPI): 141.1 x 105.8 cm; 55.56 x 41.67 inches
Original colors: 16,7 Million (24 BitsPerPixel)
Current colors: 256 (8 BitsPerPixel)
Number of unique colors: 256
Disk size: 4.48 MB (4,693,172 Bytes)

At first I was worried that the Original:Current colors would be a tip off
to anyone looking at the entropy - but they seem to sync up with a save.

[3] Reduce colors to 256 & then save the JPEG & re-open that saved JPEG.
Compression: JPEG, quality: JPEG, quality: 100, subsampling ON (2x2)
Original size: 4000 x 3000 Pixels (12.00 MPixels) (4:3)
Current size: 4000 x 3000 Pixels (12.00 MPixels) (4:3)
Print size (from DPI): 141.1 x 105.8 cm; 55.56 x 41.67 inches
Original colors: 16,7 Million (24 BitsPerPixel)
Current colors: 16,7 Million (24 BitsPerPixel)
Number of unique colors: 98371
Disk size: 9.17 MB (9,615,126 Bytes)

Whoa. That's odd. Very strange. The number of unique colors is exactly the
same for both methods but the file size is double in the JPG->256-JPG
method.

That is, the JPG->GIF->JPG resulted in exactly 98371 unique colors.
But twice the original file size.

And the JPG->256->JPG also resulted in exactly 98371 unique colors.
But the same file size as the original was.

Post by Newyana2
Yet the GIF looks OK.

The goal is for the powers that be to not realize it was done.

Whether I move the unique computer-perceptible flaws (by tilting & cropping
for example) or I paper them over (with blurs & color conversions),
whatever obfuscation techniques we use must have an end result JPG that
appears to be a normal upload (just like those from everyone else).

The only thing left now is to reduce the size as most uploads are
not the full-size image - but some reduced-size image.

[4] Reducing that last JPG->256->JPG with File->Save 80% compression alone:
Compression: JPEG, quality: 90, subsampling ON (2x2)
Original size: 4000 x 3000 Pixels (12.00 MPixels) (4:3)
Current size: 4000 x 3000 Pixels (12.00 MPixels) (4:3)
Print size (from DPI): 141.1 x 105.8 cm; 55.56 x 41.67 inches
Original colors: 16,7 Million (24 BitsPerPixel)
Current colors: 16,7 Million (24 BitsPerPixel)
Number of unique colors: 207697
Disk size: 3.13 MB (3,285,403 Bytes)

Which is interesting because we're now pretty much looking at an image
which is similar in the specs above to the original image in number of
unique colors being around 200K and the file size of around 3 MB.

Post by Newyana2
And as you noted, the color count increases when it's
converted back. Somehow the JPG conversion brings
back some sharpness.

The (JPG->GIF->JPG vs JPG-256-JPG) methods seem to have similar results.

I guess that means if I'm working on a hundred images, I'll use the batch
conversion of JPG->GIF->JPG (and batch delete the intermediate GIF).

But if I'm working on only one image, then I'll just employ the
JPG->256->JPG method (instead of creating the intermediate GIF file).

Post by Newyana2
| Why would a sharpen add so many unique colors to the image?
|
Sharpen highlights the difference along edges. If you
take a simple image -- a single color background with
a different color rectangle in the middle, say, a sharpen
will add 1 or more lies of new colors where the 2 colors
meet. In a photo you're doing that with a slightly different
hue for each pixel comparison.

Thanks for explaining why sharpen adds colors by changing the pixels along
the edges of the objects in the image. I don't know what a "resampling" is,
but most images uploaded to the Internet are likely resized where Irfanview
turns on resampling automatically with resize.

To check the effect on entropy that resizing has, I just ran this test
on that last JPG-256-JPG->90% compressed image to gauge the effect.

[5a] Resize/Resample [4] -> 800x600 (no apply resharpen)(no compression)
Compression: JPEG, quality: 100, subsampling ON (2x2)
Original size: 800 x 600 Pixels (4:3)
Current size: 800 x 600 Pixels (4:3)
Print size (from DPI): 28.2 x 21.2 cm; 11.11 x 8.33 inches
Original colors: 16,7 Million (24 BitsPerPixel)
Current colors: 16,7 Million (24 BitsPerPixel)
Number of unique colors: 70046
Disk size: 340.72 KB (348,902 Bytes)

That's a reasonable size for a typical Internet upload but let me see what
the difference would be had I done the same steps but with resampling.

[5b] Resize/Resample [4] -> 800x600 (yes apply resharpen)(no compression)
Compression: JPEG, quality: 100, subsampling ON (2x2)
Original size: 800 x 600 Pixels (4:3)
Current size: 800 x 600 Pixels (4:3)
Print size (from DPI): 28.2 x 21.2 cm; 11.11 x 8.33 inches
Original colors: 16,7 Million (24 BitsPerPixel)
Current colors: 16,7 Million (24 BitsPerPixel)
Number of unique colors: 73459
Disk size: 386.02 KB (395,283 Bytes)

Which means the "Apply sharpen after Resample" doesn't seem to change much.

BTW, it must be important to apply the sharpen after the resample because
it's the default for Irfanview.

Post by Newyana2
| But of course, I have no way of knowing if this papers over unique camera
| sensor flaws that were in the original image that moved forward
| throughout.
No. You should have virtually all new pixels, but I don't
know how their method works.

I also do NOT know how their method works (for example, if you flip an
image horizontally, does their method still work?) but from what I've
gleaned over the years, each sensor has unique flaws that they can detect.

If we assume that flaw is a "hole", for example, and if we assume that hole
(for arguments sake) is dead center in the middle of the image, then that
one pixel in the middle of the image will be "00000000" (for example).

I can move that flaw to some other spot by tilting and cropping but (unless
cropped out) that flaw would still exist with tilting and cropping alone.

I'd have to run another step to paper over that flaw entirely so that it's
no longer 00000000. What Irfanview function do you think can do that best?

Post by Newyana2
| Since the DISPLAY is a BMP, would you think it useful to paper over unique
| camera sensor flaws by snapping a screenshot and replacing it with that?
|>
A screenshot of the desktop? That should give you
just the same bitmap.

Oh no!

I didn't realize that a screenshot gets me the exact same bitmap.
Are you sure?

The reason I ask is not every screen has the same resolution as the image.
So how can it be that they're both exactly the same?

If the image has resolution of, for example, 2X the screen, then how can
teh result be the same since the screen itself can't resolve more than 1X?

I was hoping a screenshot would give me the unique flaws in my screen,
instead of the unique flaws in my camera since the resolution is different.

I should say I don't really understand resolution or DPI so I might be
confusing the two because they both seem to be the same thing to me.

Post by Newyana2
A screenshot will just send you the
byte values being sent to the graphic hardware.

This confuses me because I don't know enough about digital images to
understand it.

I'm trying to understand why a screenshot won't help by reducing uniqueness
(as it always seemed to me that it is no longer the original image).

If, for example, there's a "hole" in the center of my camera sensor such
that the middle pixel is "0000000" on the actual image, is a screenshot of
that original image (saved to a new file) also going to have that hole?

Post by Newyana2
| I think we need to fundamentally do a few things, but I'm not sure.
Maybe put it in a locked box, sseal that with wax, then
bury it under your garage floor. :)

The first step in protection is simply knowing what that first step is.
Those who don't know about this issue can't even protect against it.
At least we can try. They can't. So we're ahead of them.

Post by Newyana2
| We need to paper over the unique flaws with blurring somehow.
Are they flaws? Could they be something like particular
hues that can never show up? I don't know.

I also do not know. But I'm aware they're always writing papers on new
techniques which work better, so I think the uniqueness of the camera
sensor is what they detect - but they do so via different methods.

I think they're definitely at the point today that any image that isn't
modified by the user is definitely traced to the exact camera that took it.

That alone is a certainty. What I do not know is how much (or how little)
of a modification does one need to balance the effort against the results.

Given every picture that's uploaded has to be modified anyway (for sheer
size alone if for no other reason), it's easy enough to add more steps.

Post by Newyana2
without knowing
the method of inspection you're in the dark. But I'm guessing
your method will have changed the precise hue values of nearly
every pixel. So... pretty much a new image that just happens
to look the same to your eye.

I use a combination of steps (most of which have been mentioned but not all
of them) where it's all done in Irfanview (for convenience) and where you
have helped me a lot to understand what Irfanview can do for papering over
flaws.

One test I can try is to find a high-resolution photo on the net, and then
modify that and post it back & see, if in a few months, my modification
shows up when I do a search on the reverse image search engines.

I could do two images, one as a control (which is not modified), and
another image from the same source (which is modified).

But that wouldn't test forensics so much as how good the image search
engine is (where cropping & flipping foils many image search engines).

[Newyana2]

"Peter" <***@nospam.net> wrote

| One way we can probably check our entropy is to pick a photo from the
| Internet and then run our commands on that, & then upload it somewhere.
|

Recall that this is your project, not ours. :) I rarely take
photos and even more rarely upload them.

| > A screenshot of the desktop? That should give you
| > just the same bitmap.
|
| Oh no!
|
| I didn't realize that a screenshot gets me the exact same bitmap.
| Are you sure?
|
| The reason I ask is not every screen has the same resolution as the image.
| So how can it be that they're both exactly the same?
|

Resolution is meaningless for a DIB. I-V may say the image
is 300 dpi, for example, but that only applies when it's printed
at 300 dpi. Inches are an abstraction on a computer screen.

The typical display setting is 96 dpi, but that depends on your
screen size setting. If you have an image 200x200 and it looks,
say, 2" square on your monitor at 1920 wide, if you then change
your display to something like 800x600 then your image will be
more than 4" wide.

The graphics driver sends data for 200x200 pixels to the screen.
The image data is not changed. If you watch a movie with John
Wayne on a 17" TV or a 50" TV, how wide is John Wayne? (Sounds
like a good koan.) Regardless of the TV, the film projected to show
the movie is not changed.

| If, for example, there's a "hole" in the center of my camera sensor such
| that the middle pixel is "0000000" on the actual image, is a screenshot of
| that original image (saved to a new file) also going to have that hole?
|

Yes, of course. If you take a photo of a tree it's not going
to turn into a car just because you intended to take a photo
of a car. The whole thing is byte data. If there's a flaw causing
a black pixel then there's a black pixel.

If you take RAW photos then you can change them quite a bit
in saving to JPG because there's a lot more data there. The color
space is larger starting out. But even then, a black dot in the middle
is still part of the image. Will it be dithered out in reducing?
Probably, but I can't say for sure.

You have to get used to there being no absolute truth when
it comes to computer graphics. It's all about creating images out
of dots that each represent hues in a limited color spectrum. They're
not the real hues based on reflected light. They're an approximation
of the range of colors the human eye sees. The color sensor is biased.
For example, a bumblebee might see bright blue stripes on a magenta
flower. Maybe it won't see the magenta. I don't know. So what color
is the flower? Black with blue stripes, or magenta? The actual light
reflected cannot be fully perceived by either the bumblebee or a
human. So that's just an abstraction for practical purposes. The
color sensor in your camera will be designed to record colors in
a range that you can see. That data is then further reduced by
converting it to a numeric value in a limited range. There's no flower
in your photo. There's only a long string of bytes that represent
RGB values for dots on a grid. Depending on your monitor, display
driver, eyesight, etc, you'll see a facsimile of that flower on your
screen. Even on the same computer I see different graphics if I
boot Windows vs Linux. Yet the byte data is the same.

[Peter]

Post by Newyana2
| One way we can probably check our entropy is to pick a photo from the
| Internet and then run our commands on that, & then upload it somewhere.
|
Recall that this is your project, not ours. :) I rarely take
photos and even more rarely upload them.

Oh. No problem. I didn't mean at all to sign you up. It's my crusade.

I had meant the "we" as in a royal we (as in anyone), and not as in you and
me.

My observation was that if we (anyone) can't even fool reverse-image-search
engines, though, then we (anyone) didn't do it right because they're easy
to fool.

Post by Newyana2
|> A screenshot of the desktop? That should give you
|> just the same bitmap.
|
| I didn't realize that a screenshot gets me the exact same bitmap.
| Are you sure?
|
| The reason I ask is not every screen has the same resolution as the image.
| So how can it be that they're both exactly the same?
|
Resolution is meaningless for a DIB. I-V may say the image
is 300 dpi, for example, but that only applies when it's printed
at 300 dpi. Inches are an abstraction on a computer screen.

Thanks for letting me know that resolution is meaningless for a digital
image in terms of our (the royal our) goal of protecting against being
identified by our camera sensor pixel imperfections.

Post by Newyana2
The typical display setting is 96 dpi, but that depends on your
screen size setting. If you have an image 200x200 and it looks,
say, 2" square on your monitor at 1920 wide, if you then change
your display to something like 800x600 then your image will be
more than 4" wide.

That's what confuses me because, in that scenario you describe above, my
question is if I screenshot the whole screen (and crop to the 2" image)
when the image is 2" square versus if I screenshot the whole screen (and
crop to the 4" image), you seem to have been saying the resulting two
images are EXACTLY the same.

Are they?

It's super important to nail down the answer to that question!!!!!!
If you answer no other question ever again, that's the most important.

Post by Newyana2
The graphics driver sends data for 200x200 pixels to the screen.
The image data is not changed. If you watch a movie with John
Wayne on a 17" TV or a 50" TV, how wide is John Wayne? (Sounds
like a good koan.) Regardless of the TV, the film projected to show
the movie is not changed.

I didn't realize that the image will be the same when you do this, but can
you simply confirm with a yes or no given this scenario below.

[1] Your starting point is an image from your camera
[2] You screenshot it on Display1 and save the full-screen results
(and then you crop away the extraneous Windows blue desktop background)
[3] Same thing on Display2.

Are you saying that all three images will be exactly the same pixels?
How can that be given what is saved is always smaller in file size than the
original?

This is the most important question for us (anyone) to understand I think.

Post by Newyana2
| If, for example, there's a "hole" in the center of my camera sensor such
| that the middle pixel is "0000000" on the actual image, is a screenshot of
| that original image (saved to a new file) also going to have that hole?
Yes, of course. If you take a photo of a tree it's not going
to turn into a car just because you intended to take a photo
of a car. The whole thing is byte data. If there's a flaw causing
a black pixel then there's a black pixel.

That's not good if the screenshot reproduces the image perfectly.
But it can't because the screenshot isn't even the same file size.
Something is missing in my understanding.

Post by Newyana2
If you take RAW photos then you can change them quite a bit
in saving to JPG because there's a lot more data there. The color
space is larger starting out. But even then, a black dot in the middle
is still part of the image. Will it be dithered out in reducing?
Probably, but I can't say for sure.

If that's the case, I think tilting & cropping & flipping (if possible)
will just move the black dot. Maybe the Irfanview blurring may help paper
over the black dot.

A far deeper question (for another time) may be which blurring technique is
most effective for our (anyone's) purposes, where that's for a later
discussion as it could get too technical for me really quickly (gaussian,
bokeh, quantized, etc) as I see those options in Windows image editors.

Post by Newyana2
You have to get used to there being no absolute truth when
it comes to computer graphics. It's all about creating images out
of dots that each represent hues in a limited color spectrum. They're
not the real hues based on reflected light. They're an approximation
of the range of colors the human eye sees. The color sensor is biased.
For example, a bumblebee might see bright blue stripes on a magenta
flower. Maybe it won't see the magenta. I don't know. So what color
is the flower? Black with blue stripes, or magenta? The actual light
reflected cannot be fully perceived by either the bumblebee or a
human.

What you're trying to tell me, I think, is that the colors I see or that
the monitor sees isn't what colors are in the original image.

Do you think Irfanview "auto adjust colors" will help?

Or does that too do nothing to the saved image's pixel values?

Post by Newyana2
So that's just an abstraction for practical purposes. The
color sensor in your camera will be designed to record colors in
a range that you can see. That data is then further reduced by
converting it to a numeric value in a limited range. There's no flower
in your photo. There's only a long string of bytes that represent
RGB values for dots on a grid. Depending on your monitor, display
driver, eyesight, etc, you'll see a facsimile of that flower on your
screen. Even on the same computer I see different graphics if I
boot Windows vs Linux. Yet the byte data is the same.

Thanks for all your help as all I'm trying to do is solve the problem.

The main question to pin down the answer to is if the screenshot of the
image is the exact same pixel values, why is the screenshot a different
size than the original image?

[Newyana2]

"Peter" <***@nospam.net> wrote

| That's what confuses me because, in that scenario you describe above, my
| question is if I screenshot the whole screen (and crop to the 2" image)
| when the image is 2" square versus if I screenshot the whole screen (and
| crop to the 4" image), you seem to have been saying the resulting two
| images are EXACTLY the same.
|
| Are they?
|
| It's super important to nail down the answer to that question!!!!!!
| If you answer no other question ever again, that's the most important.
|

It's the same 200x200 image either way. Only the
display is different.


| > The graphics driver sends data for 200x200 pixels to the screen.
| > The image data is not changed. If you watch a movie with John
| > Wayne on a 17" TV or a 50" TV, how wide is John Wayne? (Sounds
| > like a good koan.) Regardless of the TV, the film projected to show
| > the movie is not changed.
|
| I didn't realize that the image will be the same when you do this, but can
| you simply confirm with a yes or no given this scenario below.
|
| [1] Your starting point is an image from your camera
| [2] You screenshot it on Display1 and save the full-screen results
| (and then you crop away the extraneous Windows blue desktop background)
| [3] Same thing on Display2.
|
| Are you saying that all three images will be exactly the same pixels?
| How can that be given what is saved is always smaller in file size than
the
| original?
|
They should be, if you save it to BMP. If you save to JPG
then you're changing the image.

| What you're trying to tell me, I think, is that the colors I see or that
| the monitor sees isn't what colors are in the original image.
|
There's no absolute color. The image saves byte
values that represent RGB. Try making a very simple BMP
and open the file in a hex editor. After the header bytes
you can see the bytes represeting color. For example, if
you save an image of pure blue then the bytes will appear
as FF 00 00 FF 00 00 FF 00 00 and so on.

Now what if you send that to a B/W TV screen? It will
be gray. If you have eye problems so that you can't see
blue then you'll perhaps see green. I don't know.

| The main question to pin down the answer to is if the screenshot of the
| image is the exact same pixel values, why is the screenshot a different
| size than the original image?

I explained that above. The screenshot is the same size
in terms of pixels. How those pixels are displayed is a different
issue. If you watch a movie on TV or in a theater it's a different
size, right? It's not different images. It's different display size.

[Peter]

Post by Newyana2
| The main question to pin down the answer to is if the screenshot of the
| image is the exact same pixel values, why is the screenshot a different
| size than the original image?
I explained that above. The screenshot is the same size
in terms of pixels. How those pixels are displayed is a different
issue. If you watch a movie on TV or in a theater it's a different
size, right? It's not different images. It's different display size.

I appreciate narrowing down the questions to the single most important one.

In my tests, a screenshot is completely different than the original image.
Loading Image...
Loading Image...
Yet you keep saying the screenshot results in the same as the original.

You know more than I do so you're likely right - but what am I doing wrong?
So I just ran this repeatable experiment which I'd ask you to test out.

1. I arbitrarily picked this image of a woodcock sitting amongst greenery.
Loading Image...

Let's never edit or re-save that image ever again.
So I made a copy called "woodcock1.jpg" instead.

Compression: JPEG, quality: 94, subsampling OFF
Original size: 4200 x 3080 Pixels (12.94 MPixels) (1.36)
Current size: 4200 x 3080 Pixels (12.94 MPixels) (1.36)
Original colors: 16,7 Million (24 BitsPerPixel)
Current colors: 16,7 Million (24 BitsPerPixel)
Number of unique colors: 329007
Disk size: 3.74 MB (3,925,976 Bytes)

2. In Irfanview, I display that woodcock1.jpg image on my first display.
In Irfanview, I adjust the window borders to about 4 inches square.
In Irfanview I fit that woodcock1.jpg image to that 4x4 inch window by
using Irfanview -> View -> Display options -> Fit images to window

This is what that looks like on the first of the two screens.
Loading Image...

Then I press the keyboard "printscreen" button.
This actually prints both screens (I don't know how to print just one).
I paste those results back over the Irfanview image.
And I crop out both of the Windows blue backgrounds.
Until I am back to just the image without the Irfanview window borders.
I save that resulting Irfanview image as woodcock2.jpg (no compression).

Compression: JPEG, quality: 100, subsampling ON (2x2)
Original size: 314 x 230 Pixels (1.36)
Current size: 314 x 230 Pixels (1.36)
Original colors: 16,7 Million (24 BitsPerPixel)
Current colors: 16,7 Million (24 BitsPerPixel)
Number of unique colors: 32347
Disk size: 73.89 KB (75,661 Bytes)

The two images (woodcock1.jpg & woodcock2.jpg) are completely different
even though the second image is just a screenshot of the first image.

The reason I'm confused by you saying the screenshot is the same bitmap as
the image was is there's no way that those two files are close in any way.

I must not be understanding what you mean then when you say that the
screenshot of the image will just get me back to the bits in the image.

Why are my screenshot results completely different from the original then?
https://appvoices.org/images/uploads/2016/04/woodcock.jpg
https://i.postimg.cc/05J1xDrG/woodcock1.jpg
https://i.postimg.cc/4yNbvf6w/fourinchesonscreen.jpg
https://i.postimg.cc/vHFFmT6J/woodcock2.jpg

[Newyana2]

"Peter" <***@nospam.net> wrote

| Why are my screenshot results completely different from the original then?
| https://appvoices.org/images/uploads/2016/04/woodcock.jpg
| https://i.postimg.cc/05J1xDrG/woodcock1.jpg
| https://i.postimg.cc/4yNbvf6w/fourinchesonscreen.jpg
| https://i.postimg.cc/vHFFmT6J/woodcock2.jpg

As I said, if you save to JPG it will be different. Even
zero compression still compresses and will involve data loss.
But isn't this a sidetrack? It's not relevant to use
screenshots. The point is just to find any way to change
the pixels, then find software to test whether it worked.

[Peter]

Post by Newyana2
| Why are my screenshot results completely different from the original then?
| https://appvoices.org/images/uploads/2016/04/woodcock.jpg
| https://i.postimg.cc/05J1xDrG/woodcock1.jpg
| https://i.postimg.cc/4yNbvf6w/fourinchesonscreen.jpg
| https://i.postimg.cc/vHFFmT6J/woodcock2.jpg
As I said, if you save to JPG it will be different.]

Thanks for trying to understand, where I'm fully aware that a JPG is lossy
compression so that every save set at less than no compression, loses bits.

But I wasn't aware that a save with no compression also loses bits.

But that's a lot of bits to lose in a single save with no compression!
https://i.postimg.cc/05J1xDrG/woodcock1.jpg = Disk size: 3.74 MB
https://i.postimg.cc/vHFFmT6J/woodcock2.jpg = Disk size: 73.89 KB
Especially as the only thing in between was a single screenshot & crop.

Post by Newyana2
Even zero compression still compresses and will involve data loss.

Because it is so easy to do, and seemingly it increases entropy a lot,
the important question to answer is what does a screenshot actually do.

If a screenshot is so faithful to the bits, why is simply running a single
screenshot & cropping back to the image & saving once losing so many bits?

Even the number of unique colors is hugely different in the screenshot.
woodcock1.jpg = Disk size: 3.74 MB, 329007 unique colors
woodcock2.jpg = Disk size: 73.89 KB, 32347 unique colors

Post by Newyana2
But isn't this a sidetrack? It's not relevant to use screenshots.

It may be wrong to use a screenshot and then it's irrelevant, but using a
screenshot is admittedly very easy and it drastically changes the image.

The image is so drastically changed that it could maybe be the best method.

Post by Newyana2
The point is just to find any way to change
the pixels, then find software to test whether it worked.

Today I use a combination of methods (most mentioned already) to move all
sensor flaws to both a different X location and to a different Y location.

But of course, moving the flaw (for example, flipping or rotating) will not
change the relationship between any two flaws so papering over is needed.

I use a variety of methods to paper over flaws (blur & sharpen & reduce
colors for example - which is what sparked this conversation initially).

But what's most important to narrow down the answer to (because it seems so
functionally powerful and it's very easy to do) is what a screenshot does.

If a screenshot is so faithful to the original, why can't I reproduce that?

[Paul]

Post by Peter

Post by Newyana2
| Why are my screenshot results completely different from the original then?
| https://appvoices.org/images/uploads/2016/04/woodcock.jpg
| https://i.postimg.cc/05J1xDrG/woodcock1.jpg
| https://i.postimg.cc/4yNbvf6w/fourinchesonscreen.jpg
| https://i.postimg.cc/vHFFmT6J/woodcock2.jpg
As I said, if you save to JPG it will be different.]

Thanks for trying to understand, where I'm fully aware that a JPG is lossy
compression so that every save set at less than no compression, loses bits.
But I wasn't aware that a save with no compression also loses bits.
But that's a lot of bits to lose in a single save with no compression!
https://i.postimg.cc/05J1xDrG/woodcock1.jpg = Disk size: 3.74 MB
https://i.postimg.cc/vHFFmT6J/woodcock2.jpg = Disk size: 73.89 KB
Especially as the only thing in between was a single screenshot & crop.

Post by Newyana2
Even zero compression still compresses and will involve data loss.

Because it is so easy to do, and seemingly it increases entropy a lot,
the important question to answer is what does a screenshot actually do.
If a screenshot is so faithful to the bits, why is simply running a single
screenshot & cropping back to the image & saving once losing so many bits?
Even the number of unique colors is hugely different in the screenshot.
woodcock1.jpg = Disk size: 3.74 MB, 329007 unique colors
woodcock2.jpg = Disk size: 73.89 KB, 32347 unique colors

Post by Newyana2
But isn't this a sidetrack? It's not relevant to use screenshots.

It may be wrong to use a screenshot and then it's irrelevant, but using a
screenshot is admittedly very easy and it drastically changes the image.
The image is so drastically changed that it could maybe be the best method.

Post by Newyana2
The point is just to find any way to change
the pixels, then find software to test whether it worked.

Today I use a combination of methods (most mentioned already) to move all
sensor flaws to both a different X location and to a different Y location.
But of course, moving the flaw (for example, flipping or rotating) will not
change the relationship between any two flaws so papering over is needed.
I use a variety of methods to paper over flaws (blur & sharpen & reduce
colors for example - which is what sparked this conversation initially).
But what's most important to narrow down the answer to (because it seems so
functionally powerful and it's very easy to do) is what a screenshot does.
If a screenshot is so faithful to the original, why can't I reproduce that?

The original image is 4200x3080.

To present that on a 1920x1080 screen, means processing
a clump of pixels and "summarizing" that content, with
a new synthetic pixel. Depending on how many pixels are
being averaged, that's going to "damp" the camera sensor
signature a bit.

Paul

[Peter]

Post by Paul

Post by Peter
If a screenshot is so faithful to the original, why can't I reproduce that?

The original image is 4200x3080.
To present that on a 1920x1080 screen, means processing
a clump of pixels and "summarizing" that content, with
a new synthetic pixel. Depending on how many pixels are
being averaged, that's going to "damp" the camera sensor
signature a bit.

That's why I didn't understand what was being said when it was said that a
screenshot is exactly the same pixels. They're not even close in my tests.

While that's confusing because what people said who know more than I do was
that the screenshot is exactly the same, if it does "summarize the
content", that's an almost perfect way to get rid of explicit flaws, I
would think.

Do you see where I'm going?

If the goal is to get rid of specific camera sensor flaws (which we have no
idea what they are), then it seems that "processing a clump of pixels" as
if they were a single pixel, is an extremely useful step for that purpose.

Especially as it take no more effort or special tools than the pressing of
a keyboard button and saving the results.

Does my logic make sense to you given the goal is obfuscating sensor flaws?
Or am I missing something?

[Paul]

Post by Peter

Post by Paul

Post by Peter
If a screenshot is so faithful to the original, why can't I reproduce that?

The original image is 4200x3080.
To present that on a 1920x1080 screen, means processing
a clump of pixels and "summarizing" that content, with
a new synthetic pixel. Depending on how many pixels are
being averaged, that's going to "damp" the camera sensor
signature a bit.

That's why I didn't understand what was being said when it was said that a
screenshot is exactly the same pixels. They're not even close in my tests.
While that's confusing because what people said who know more than I do was
that the screenshot is exactly the same, if it does "summarize the
content", that's an almost perfect way to get rid of explicit flaws, I
would think.
Do you see where I'm going?
If the goal is to get rid of specific camera sensor flaws (which we have no
idea what they are), then it seems that "processing a clump of pixels" as
if they were a single pixel, is an extremely useful step for that purpose.
Especially as it take no more effort or special tools than the pressing of
a keyboard button and saving the results.
Does my logic make sense to you given the goal is obfuscating sensor flaws?
Or am I missing something?

It attenuates the amplitude of the information.

A question would be, is the effect different to
start with, with a sensor having fewer pixels (larger area) ?
Or a sensor manufactured a different way ?
Like say, CCD versus CMOS, or the cheap kind used
in webcams, versus other camera types ?

Paul

[Nic]

Post by Paul

Post by Peter

Post by Newyana2
| Why are my screenshot results completely different from the original then?
| https://appvoices.org/images/uploads/2016/04/woodcock.jpg
| https://i.postimg.cc/05J1xDrG/woodcock1.jpg
| https://i.postimg.cc/4yNbvf6w/fourinchesonscreen.jpg
| https://i.postimg.cc/vHFFmT6J/woodcock2.jpg
As I said, if you save to JPG it will be different.]

Thanks for trying to understand, where I'm fully aware that a JPG is lossy
compression so that every save set at less than no compression, loses bits.
But I wasn't aware that a save with no compression also loses bits.
But that's a lot of bits to lose in a single save with no compression!
https://i.postimg.cc/05J1xDrG/woodcock1.jpg = Disk size: 3.74 MB
https://i.postimg.cc/vHFFmT6J/woodcock2.jpg = Disk size: 73.89 KB
Especially as the only thing in between was a single screenshot & crop.

Post by Newyana2
Even zero compression still compresses and will involve data loss.

Because it is so easy to do, and seemingly it increases entropy a lot,
the important question to answer is what does a screenshot actually do.
If a screenshot is so faithful to the bits, why is simply running a single
screenshot & cropping back to the image & saving once losing so many bits?
Even the number of unique colors is hugely different in the screenshot.
woodcock1.jpg = Disk size: 3.74 MB, 329007 unique colors
woodcock2.jpg = Disk size: 73.89 KB, 32347 unique colors

Post by Newyana2
But isn't this a sidetrack? It's not relevant to use screenshots.

It may be wrong to use a screenshot and then it's irrelevant, but using a
screenshot is admittedly very easy and it drastically changes the image.
The image is so drastically changed that it could maybe be the best method.

Post by Newyana2
The point is just to find any way to change
the pixels, then find software to test whether it worked.

Today I use a combination of methods (most mentioned already) to move all
sensor flaws to both a different X location and to a different Y location.
But of course, moving the flaw (for example, flipping or rotating) will not
change the relationship between any two flaws so papering over is needed.
I use a variety of methods to paper over flaws (blur & sharpen & reduce
colors for example - which is what sparked this conversation initially).
But what's most important to narrow down the answer to (because it seems so
functionally powerful and it's very easy to do) is what a screenshot does.
If a screenshot is so faithful to the original, why can't I reproduce that?

The original image is 4200x3080.
To present that on a 1920x1080 screen, means processing
a clump of pixels and "summarizing" that content, with
a new synthetic pixel. Depending on how many pixels are
being averaged, that's going to "damp" the camera sensor
signature a bit.
Paul

What happens when Grayscale is used, from camera to jpg or bmp?

[Paul]

Post by Nic
What happens when Grayscale is used, from camera to jpg or bmp?

JPG processing, tends to spray into the colorspace.

You can study that, by "counting colors" with Irfanview,
and comparing the number of colors in the source picture,
versus the number in the output.

BMP is transparent, up to the limits of its 24-bit carriage.
Even with indexed color, you can in some of the cases,
convert that to 24 bit, without affecting anything.
Modern hardware, can create more bits than you know
what to do with.

PNG has more options for carriage, and can shrink some
special-case source images, significantly.

There is a tendency to abuse JPG, and use it for things
it isn't optimal for. For photography, it's sins may be
a worthwhile tradeoff, if storage space somewhere is at
a premium. Making it compress cartoon cells, isn't a good
usage of it. (Cartoons might work better as a GIF.)

Paul

[rocco portelli]

Post by Peter
Because it is so easy to do, and seemingly it increases entropy a lot,
the important question to answer is what does a screenshot actually do.

I wonder if that screenshot & crop is the same as a resize to 4x4 inches?

[Paul]

Post by Newyana2
| In Irfanview, you can Decrease Color Depth to any level but how can you
| either increase it or figure out what the current color depth is set to?
If you haven't changed anything then you'll mostly know
by file type. JPGs will be 24-bit, GIFs are 8-bit, PNGs
32-bit. 24-bit is actually the max in standard usage.
The extra 8 bits in PNGs are for transparency values.
BMPs can be 2-bit to 24-bit, but most people don't
see BMPs much these days.
Do you know what color depth is? A monitor these days
displays 24-bit color, which means 256 different hues of
red, green and blue. 0-0-0 is black. 255-255-255 is white.
(Unless you're on a Mac, in which case I think it's 18-bit
color. Which means it can only display 64 color gradients for
R, G and B -- it's missing 192 hues out of 256, so it dithers
pixels to the nearest color.)
Raster image file formats store those values as numbers
representing a pixel grid. They're all bitmap when they
display, but each format stores the data differently.
Why would you want to change color depth? If you want
to do something like save a JPG as GIF then you'll lose a lot
of the color data. If you do the reverse you won't get more
colors. The original GIF color table entries are all that you'll
see unless you then edit the image in 24-bit.

JPG is subject to rounding errors. These spray into the
color space.

For example, open a GIF (8 bit, indexed), now save as JPG,
then open in Irfanview. Do "Information". It counts the colors
for you. There could be 10,000 colors in there. Now, if
you plot the "locus" of the colors, you'll find colors very
close to the main 256 colors, all closely clustered around their
parent dot.

Now, ask a tool to do a color space reduction. Ask to go from
24 bit RGB (full color space) to 8-bit indexed. All of the
colors which are nearest their parent dot, are converted to
the parent dot value. Now, the image has 256 colors again,
and they're indexes into a 24 bit color table.

If you then ask to save the image as a GIF, the save should go
very fast, because the colorspace is now pretty close to what
GIF wants.

*******

Increasing the color space, that operation "does nothing".
What started as a 0xaa 0xaa 0xaa pixel remains the same color.
If you change to an HDR (High Dynamic Range) pixel with
(3) ten bit values, then the lower two bits will be 00.

It's when you do subsequent math on the representation, that
some of the bit values may move around. Depending on what
you're doing.

*******

When you save to PNG, one of the options is to control
the number of bits used. You can use 8 bits per color.
You can use 2 bits per color. This is all part of the zillion
compression options when making a PNG.

And none of these transformations are of particular interest
to rec.photo.digital , as they're struggling to preserve what
they've got. Only web people or cartoon people, revel in
color space transformations for size reduction or other
purposes. If you do color space reductions, you can get
banding on the screen when you look at the result. A photographer
does not want banding on a screen, or in print.

What I can see of Irfanview, it does not look like you
can raise the color space, higher than the screen capability.
My screen isn't HDR10, so the color space apparently
has no need to go above 24-bit RGB.

The versions of Photoshop I've got, from years ago, their
internal representation adjusts for the expected dynamics
of any math. When you average two pictures, (A+B)/2,
it uses 9 bits per pixel when doing the math. Averaging
two pictures, is a way of reducing sensor noise. And it
only works, if you shoot two identical picture with a
tripod, and nothing in the scene is moving. I made a user
manual once, with static photos, and used that technique
to clean up the poor sensor noise. The scene was illuminated,
but the sensor used was a joke. But you do the best with
what you've got.

There are a number of other wasteful tools, that use way too
many bits when doing math (like, three floats), and these
serve to gobble down RAM when processing images. The approach
Photoshop was using at the time, was pretty optimal.

Paul