Resolution - Benefits of higher Megapixel - effects of jpg compression

I've used an Olympus C-2100UZ for about four years extremely happily.
Since it's "only" a 2mp camera I recently decided to think about
upgrading. Before doing so I decided to see if I'm getting the most
out of my current camera. I realized that I've been mostly shooting
in HQ mode.

I performed a series of tests shooting in both HQ (0.45MB average) and
SHQ (1.3MB average) mode, both handheld and with a tripod. Despite
the nearly 3x difference in file size I could find no realistic
difference between the results of the two files either blown up on
screen or when printed. BIG SURPRISE.

I did a search and found the following thread:
http://groups.google.com/groups?hl=e...6btnG%3DSearch
What I concluded from its discussion is that the only person who seems
to have actually tested found the same thing I did.

I also looked at the information that PaintShop Pro provides on these
images. It lists them each as as 1600x1200, no difference between
them.

So.... My questions
#1) Does anybody have an explanation why a filesize three times
greater doesn't produce a noticeably different image when viewed
onscreen or printed. My image was of a brick wall with foliage to
have lots of both straight lines and organic shapes.

#2) If I upgrade to a 4mp or higher camera of current vintage (e.g.
Pansonic FZ-20) will I find the same lack of difference between
Panasonic's version of HQ and SHQ?

#3) With that same upgrade, essentially doubling the number of
pixels, am I correct in expecting a doubling of the number of pixels
in the image and thus a resultant increase in resolution of 1.4 on
each axis.

Many thanks

Jim Mitchell

"Jim Mitchell" wrote in message
om...
...
I performed a series of tests shooting in both HQ (0.45MB average) and
SHQ (1.3MB average) mode, both handheld and with a tripod. Despite
the nearly 3x difference in file size I could find no realistic
difference between the results of the two files either blown up on
screen or when printed. BIG SURPRISE.

I did the same test with my 3mp Canon S30, comparing "superfine",
"fine" and "normal" modes. There was a difference between
between the modes. If I blew the photos up to the point where
I could see each individual pixel, I could tell that the colors
had changed been the highest quality and next highest quality
modes. But I could not tell the difference simply by looking at
the photos at a normal 72 dpi screen resolution. I could tell a
small difference between the middle and low quality modes, so
I settled on using the middle quality mode.

...
I also looked at the information that PaintShop Pro provides on these
images. It lists them each as as 1600x1200, no difference between
them.

Right. Compression does not change the number of pixels,
it changes the color of pixels. The idea is to reduce the number
of color differences in each individual cluster of pixels - enabling
the program to use fewer bits to store those differences when
compressed. When uncompressed, the number of pixels stays
the same.

If compression is too great, "artifacts" are created. What I think
these are are places where compression moved colors away
from each other, producing apparent objects in the image that
weren't in the original. You will see these if you use very low
quality settings in your PSP Save As settings.

So.... My questions
#1) Does anybody have an explanation why a filesize three times
greater doesn't produce a noticeably different image when viewed
onscreen or printed. My image was of a brick wall with foliage to
have lots of both straight lines and organic shapes.

Yes. The explanation is that the eye does not perceive tiny
differences in color - especially if the overall brightness stays
the same. JPEG compression uses that fact about human eye
and brain color processing to achieve high compression by
reducing adjacent and nearly adjacent color distinctions to a
smaller number of distinctions. Fine color differences are lost,
but the eye and brain don't notice.

As an analogy, think about how film or television works. In
film, 24 still images per second are projected onto a screen,
but the eye and brain put them together, without flicker, into
a continuously moving image. To a differently constructed
observer, these images might appear to be 24 flickering still
images instead of one smooth motion image.

So too, to a differently constructed observer, JPEG compression
might be obvious. But it is not obvious to humans.

#2) If I upgrade to a 4mp or higher camera of current vintage (e.g.
Pansonic FZ-20) will I find the same lack of difference between
Panasonic's version of HQ and SHQ?

Every manufacturer uses different settings for compression levels.
Panasonic's may or may not be very similar to Olympus.

#3) With that same upgrade, essentially doubling the number of
pixels, am I correct in expecting a doubling of the number of pixels
in the image and thus a resultant increase in resolution of 1.4 on
each axis.

Yes. You can expect 1.414 increase in resolution on each axis.

There are interesting issues concerning whether this increases
apparent sharpness.

I think what you'll find is that in some images, the eye is not
concentrating on fine detail. For those images, the apparent
sharpness difference between the 2 and 4 mp images will not
be very great.

For others, the eye will be focussed tightly on detail and the
apparent sharpness difference will be very acute - with the
2 mp image looking terrible by comparison with the 4.

If you make prints, there will also be issues concerning
the print size. At 3x5, and maybe 4x6, you may not be
able to tell the difference between 2 and 4 mp. As you go
to larger prints, the difference will become more and more
obvious. At 8x10 it will be very apparent for many kinds
of images - though still not so apparent for some of them.

Alan

"Jim Mitchell" wrote in message
om...
...
I performed a series of tests shooting in both HQ (0.45MB average) and
SHQ (1.3MB average) mode, both handheld and with a tripod. Despite
the nearly 3x difference in file size I could find no realistic
difference between the results of the two files either blown up on
screen or when printed. BIG SURPRISE.

I did the same test with my 3mp Canon S30, comparing "superfine",
"fine" and "normal" modes. There was a difference between
between the modes. If I blew the photos up to the point where
I could see each individual pixel, I could tell that the colors
had changed been the highest quality and next highest quality
modes. But I could not tell the difference simply by looking at
the photos at a normal 72 dpi screen resolution. I could tell a
small difference between the middle and low quality modes, so
I settled on using the middle quality mode.

...
I also looked at the information that PaintShop Pro provides on these
images. It lists them each as as 1600x1200, no difference between
them.

Right. Compression does not change the number of pixels,
it changes the color of pixels. The idea is to reduce the number
of color differences in each individual cluster of pixels - enabling
the program to use fewer bits to store those differences when
compressed. When uncompressed, the number of pixels stays
the same.

If compression is too great, "artifacts" are created. What I think
these are are places where compression moved colors away
from each other, producing apparent objects in the image that
weren't in the original. You will see these if you use very low
quality settings in your PSP Save As settings.

So.... My questions
#1) Does anybody have an explanation why a filesize three times
greater doesn't produce a noticeably different image when viewed
onscreen or printed. My image was of a brick wall with foliage to
have lots of both straight lines and organic shapes.

Yes. The explanation is that the eye does not perceive tiny
differences in color - especially if the overall brightness stays
the same. JPEG compression uses that fact about human eye
and brain color processing to achieve high compression by
reducing adjacent and nearly adjacent color distinctions to a
smaller number of distinctions. Fine color differences are lost,
but the eye and brain don't notice.

As an analogy, think about how film or television works. In
film, 24 still images per second are projected onto a screen,
but the eye and brain put them together, without flicker, into
a continuously moving image. To a differently constructed
observer, these images might appear to be 24 flickering still
images instead of one smooth motion image.

So too, to a differently constructed observer, JPEG compression
might be obvious. But it is not obvious to humans.

#2) If I upgrade to a 4mp or higher camera of current vintage (e.g.
Pansonic FZ-20) will I find the same lack of difference between
Panasonic's version of HQ and SHQ?

Every manufacturer uses different settings for compression levels.
Panasonic's may or may not be very similar to Olympus.

#3) With that same upgrade, essentially doubling the number of
pixels, am I correct in expecting a doubling of the number of pixels
in the image and thus a resultant increase in resolution of 1.4 on
each axis.

Yes. You can expect 1.414 increase in resolution on each axis.

There are interesting issues concerning whether this increases
apparent sharpness.

I think what you'll find is that in some images, the eye is not
concentrating on fine detail. For those images, the apparent
sharpness difference between the 2 and 4 mp images will not
be very great.

For others, the eye will be focussed tightly on detail and the
apparent sharpness difference will be very acute - with the
2 mp image looking terrible by comparison with the 4.

If you make prints, there will also be issues concerning
the print size. At 3x5, and maybe 4x6, you may not be
able to tell the difference between 2 and 4 mp. As you go
to larger prints, the difference will become more and more
obvious. At 8x10 it will be very apparent for many kinds
of images - though still not so apparent for some of them.

Alan

"Jim Mitchell" wrote in message
om...
...
I performed a series of tests shooting in both HQ (0.45MB average) and
SHQ (1.3MB average) mode, both handheld and with a tripod. Despite
the nearly 3x difference in file size I could find no realistic
difference between the results of the two files either blown up on
screen or when printed. BIG SURPRISE.

I did the same test with my 3mp Canon S30, comparing "superfine",
"fine" and "normal" modes. There was a difference between
between the modes. If I blew the photos up to the point where
I could see each individual pixel, I could tell that the colors
had changed been the highest quality and next highest quality
modes. But I could not tell the difference simply by looking at
the photos at a normal 72 dpi screen resolution. I could tell a
small difference between the middle and low quality modes, so
I settled on using the middle quality mode.

...
I also looked at the information that PaintShop Pro provides on these
images. It lists them each as as 1600x1200, no difference between
them.

Right. Compression does not change the number of pixels,
it changes the color of pixels. The idea is to reduce the number
of color differences in each individual cluster of pixels - enabling
the program to use fewer bits to store those differences when
compressed. When uncompressed, the number of pixels stays
the same.

If compression is too great, "artifacts" are created. What I think
these are are places where compression moved colors away
from each other, producing apparent objects in the image that
weren't in the original. You will see these if you use very low
quality settings in your PSP Save As settings.

So.... My questions
#1) Does anybody have an explanation why a filesize three times
greater doesn't produce a noticeably different image when viewed
onscreen or printed. My image was of a brick wall with foliage to
have lots of both straight lines and organic shapes.

Yes. The explanation is that the eye does not perceive tiny
differences in color - especially if the overall brightness stays
the same. JPEG compression uses that fact about human eye
and brain color processing to achieve high compression by
reducing adjacent and nearly adjacent color distinctions to a
smaller number of distinctions. Fine color differences are lost,
but the eye and brain don't notice.

As an analogy, think about how film or television works. In
film, 24 still images per second are projected onto a screen,
but the eye and brain put them together, without flicker, into
a continuously moving image. To a differently constructed
observer, these images might appear to be 24 flickering still
images instead of one smooth motion image.

So too, to a differently constructed observer, JPEG compression
might be obvious. But it is not obvious to humans.

#2) If I upgrade to a 4mp or higher camera of current vintage (e.g.
Pansonic FZ-20) will I find the same lack of difference between
Panasonic's version of HQ and SHQ?

Every manufacturer uses different settings for compression levels.
Panasonic's may or may not be very similar to Olympus.

#3) With that same upgrade, essentially doubling the number of
pixels, am I correct in expecting a doubling of the number of pixels
in the image and thus a resultant increase in resolution of 1.4 on
each axis.

Yes. You can expect 1.414 increase in resolution on each axis.

There are interesting issues concerning whether this increases
apparent sharpness.

I think what you'll find is that in some images, the eye is not
concentrating on fine detail. For those images, the apparent
sharpness difference between the 2 and 4 mp images will not
be very great.

For others, the eye will be focussed tightly on detail and the
apparent sharpness difference will be very acute - with the
2 mp image looking terrible by comparison with the 4.

If you make prints, there will also be issues concerning
the print size. At 3x5, and maybe 4x6, you may not be
able to tell the difference between 2 and 4 mp. As you go
to larger prints, the difference will become more and more
obvious. At 8x10 it will be very apparent for many kinds
of images - though still not so apparent for some of them.

Alan

On 12 Sep 2004 12:56:18 -0700, (Jim Mitchell) wrote:


So.... My questions
#1) Does anybody have an explanation why a filesize three times
greater doesn't produce a noticeably different image when viewed
onscreen or printed. My image was of a brick wall with foliage to
have lots of both straight lines and organic shapes.

A lot depends on the software that did the compression, and just what you look
for when comparing. I've found that I have to compare at x8 magnification to see
differences in 'fine' and 'best' etc.

The monitor screen will never show any difference at x1 because it's resolution
is so low. As for printing - sometimes the printers software can fix a poor
photo as well.


#2) If I upgrade to a 4mp or higher camera of current vintage (e.g.
Pansonic FZ-20) will I find the same lack of difference between
Panasonic's version of HQ and SHQ?

Probably... but you will find a big difference in the bigger resolution photo.


#3) With that same upgrade, essentially doubling the number of
pixels, am I correct in expecting a doubling of the number of pixels
in the image and thus a resultant increase in resolution of 1.4 on
each axis.

Yes of course you go by area... you need 4 times the pixels to double each
dimension.

FYI... I once did a lot of testing of a Kodak 280 in all modes. I found that a
small picture at high quality had the same file size as a big picture at regular
quality - but the big picture looked far better! The reason? When you blow up
the small pic to equal the larger one, all the pixels get doubled and blurred
and there goes the quality - right out the window! So my rule is to keep the
highest size but cut the compression quality when I want to save memory card
space.

I haven't done too much testing on my D70, but I leave it in regular quality for
most stuff I shoot. Since a standard 300 DPI print would be about 8x10 I would
never see a problem unless I could print it bigger!


Many thanks

Jim Mitchell

Jim Mitchell wrote:
I've used an Olympus C-2100UZ for about four years extremely happily.
Since it's "only" a 2mp camera I recently decided to think about
upgrading. Before doing so I decided to see if I'm getting the most
out of my current camera. I realized that I've been mostly shooting
in HQ mode.

I performed a series of tests shooting in both HQ (0.45MB average) and
SHQ (1.3MB average) mode, both handheld and with a tripod. Despite
the nearly 3x difference in file size I could find no realistic
difference between the results of the two files either blown up on
screen or when printed. BIG SURPRISE.

I did a search and found the following thread:
http://groups.google.com/groups?hl=e...6btnG%3DSearch
What I concluded from its discussion is that the only person who seems
to have actually tested found the same thing I did.

I also looked at the information that PaintShop Pro provides on these
images. It lists them each as as 1600x1200, no difference between
them.

So.... My questions
#1) Does anybody have an explanation why a filesize three times
greater doesn't produce a noticeably different image when viewed
onscreen or printed. My image was of a brick wall with foliage to
have lots of both straight lines and organic shapes.

#2) If I upgrade to a 4mp or higher camera of current vintage (e.g.
Pansonic FZ-20) will I find the same lack of difference between
Panasonic's version of HQ and SHQ?

#3) With that same upgrade, essentially doubling the number of
pixels, am I correct in expecting a doubling of the number of pixels
in the image and thus a resultant increase in resolution of 1.4 on
each axis.

Many thanks

Jim Mitchell
I've read the same thing, that you can't tell the differece between
jpg's made from the same number of pixels, but with different
compression ratios. So, I leave my C5050 set on the 5 mpixel mode,
but with the highest compression jpeg output. Now I suspect that
if you wanted to do a lot of post editing with photo studio (such
as for a contest submission) you might want to store the image as a
low compression jpeg, a tiff, or even in raw mode. But for everyday
(family shots, vacation,etc) use, the high compression jpeg is fine.

Jim Mitchell wrote:
I've used an Olympus C-2100UZ for about four years extremely happily.
Since it's "only" a 2mp camera I recently decided to think about
upgrading. Before doing so I decided to see if I'm getting the most
out of my current camera. I realized that I've been mostly shooting
in HQ mode.

I performed a series of tests shooting in both HQ (0.45MB average) and
SHQ (1.3MB average) mode, both handheld and with a tripod. Despite
the nearly 3x difference in file size I could find no realistic
difference between the results of the two files either blown up on
screen or when printed. BIG SURPRISE.

I did a search and found the following thread:
http://groups.google.com/groups?hl=e...6btnG%3DSearch
What I concluded from its discussion is that the only person who seems
to have actually tested found the same thing I did.

I also looked at the information that PaintShop Pro provides on these
images. It lists them each as as 1600x1200, no difference between
them.

So.... My questions
#1) Does anybody have an explanation why a filesize three times
greater doesn't produce a noticeably different image when viewed
onscreen or printed. My image was of a brick wall with foliage to
have lots of both straight lines and organic shapes.

#2) If I upgrade to a 4mp or higher camera of current vintage (e.g.
Pansonic FZ-20) will I find the same lack of difference between
Panasonic's version of HQ and SHQ?

#3) With that same upgrade, essentially doubling the number of
pixels, am I correct in expecting a doubling of the number of pixels
in the image and thus a resultant increase in resolution of 1.4 on
each axis.

Many thanks

Jim Mitchell
I've read the same thing, that you can't tell the differece between
jpg's made from the same number of pixels, but with different
compression ratios. So, I leave my C5050 set on the 5 mpixel mode,
but with the highest compression jpeg output. Now I suspect that
if you wanted to do a lot of post editing with photo studio (such
as for a contest submission) you might want to store the image as a
low compression jpeg, a tiff, or even in raw mode. But for everyday
(family shots, vacation,etc) use, the high compression jpeg is fine.

(Jim Mitchell) wrote in
om:

#1) Does anybody have an explanation why a filesize three times
greater doesn't produce a noticeably different image when viewed
onscreen or printed. My image was of a brick wall with foliage to
have lots of both straight lines and organic shapes.


The places where the compression will be most evident are places with high
contrast and fine detail.

A good real world example is bare tree brances in the winter against the
sky. You might be able to simulate that with a page of newsprint. Needles
in a pine tree are another good real world example.

The example you give, bricks with leaves, sounds fairly "safe" for hiding
..jpg artifacts.


Bob


--
Delete the inverse SPAM to reply

In message , Jim
Mitchell writes
I've used an Olympus C-2100UZ for about four years extremely happily.
Since it's "only" a 2mp camera I recently decided to think about
upgrading. Before doing so I decided to see if I'm getting the most
out of my current camera. I realized that I've been mostly shooting
in HQ mode.

I performed a series of tests shooting in both HQ (0.45MB average) and
SHQ (1.3MB average) mode, both handheld and with a tripod. Despite
the nearly 3x difference in file size I could find no realistic
difference between the results of the two files either blown up on
screen or when printed. BIG SURPRISE.

You are probably not looking carefully enough and/or using the right
sorts of targets to see any of the JPEG artefacts that result. One very
simple example that is very tough on JPEG is a shot of a dark power line
diagonally against clear blue sky.

There is nothing else to distract the eye so when zoomed on the screen
display or printed the artefacts are more easily visible.

NB It is fairly unusual for there to be much obvious visual difference
between a first generation SHQ JPEG and a RAW file.

I did a search and found the following thread:
http://groups.google.com/groups?hl=e...safe=off&threa
dm=nut10t046d3vj7aslhs9bjs5k8sqmlvkd7%404ax.com&r num=1&prev=/groups%3Fnu
m%3D100%26hl%3Den%26lr%3Dlang_en%26ie%3DUTF-8%26safe%3Doff%26q%3Djpg%2BH
Q%2BSHQ%2BOlympus%2Bc2100%26btnG%3DSearch
What I concluded from its discussion is that the only person who seems
to have actually tested found the same thing I did.

I suggest you look for one of the past threads here where a blind test
challenge was issued. Only a handful of practitioners correctly called
the ordering of the higher quality JPEG compression images.

So.... My questions
#1) Does anybody have an explanation why a filesize three times
greater doesn't produce a noticeably different image when viewed
onscreen or printed. My image was of a brick wall with foliage to
have lots of both straight lines and organic shapes.

But probably not enough contrast and too much distracting detail to see
any artefacts at all. They will be there but you cannot see them. JPEG
is a very cunning method - the errors made in frequency space do not
show up obviously in the image for modest levels of compression.

#2) If I upgrade to a 4mp or higher camera of current vintage (e.g.
Pansonic FZ-20) will I find the same lack of difference between
Panasonic's version of HQ and SHQ?

Yes. But that is down to your testing.

Regards,
--
Martin Brown

In message , Jim
Mitchell writes
I've used an Olympus C-2100UZ for about four years extremely happily.
Since it's "only" a 2mp camera I recently decided to think about
upgrading. Before doing so I decided to see if I'm getting the most
out of my current camera. I realized that I've been mostly shooting
in HQ mode.

I performed a series of tests shooting in both HQ (0.45MB average) and
SHQ (1.3MB average) mode, both handheld and with a tripod. Despite
the nearly 3x difference in file size I could find no realistic
difference between the results of the two files either blown up on
screen or when printed. BIG SURPRISE.

You are probably not looking carefully enough and/or using the right
sorts of targets to see any of the JPEG artefacts that result. One very
simple example that is very tough on JPEG is a shot of a dark power line
diagonally against clear blue sky.

There is nothing else to distract the eye so when zoomed on the screen
display or printed the artefacts are more easily visible.

NB It is fairly unusual for there to be much obvious visual difference
between a first generation SHQ JPEG and a RAW file.

I did a search and found the following thread:
http://groups.google.com/groups?hl=e...safe=off&threa
dm=nut10t046d3vj7aslhs9bjs5k8sqmlvkd7%404ax.com&r num=1&prev=/groups%3Fnu
m%3D100%26hl%3Den%26lr%3Dlang_en%26ie%3DUTF-8%26safe%3Doff%26q%3Djpg%2BH
Q%2BSHQ%2BOlympus%2Bc2100%26btnG%3DSearch
What I concluded from its discussion is that the only person who seems
to have actually tested found the same thing I did.

I suggest you look for one of the past threads here where a blind test
challenge was issued. Only a handful of practitioners correctly called
the ordering of the higher quality JPEG compression images.

So.... My questions
#1) Does anybody have an explanation why a filesize three times
greater doesn't produce a noticeably different image when viewed
onscreen or printed. My image was of a brick wall with foliage to
have lots of both straight lines and organic shapes.

But probably not enough contrast and too much distracting detail to see
any artefacts at all. They will be there but you cannot see them. JPEG
is a very cunning method - the errors made in frequency space do not
show up obviously in the image for modest levels of compression.

#2) If I upgrade to a 4mp or higher camera of current vintage (e.g.
Pansonic FZ-20) will I find the same lack of difference between
Panasonic's version of HQ and SHQ?

Yes. But that is down to your testing.

Regards,
--
Martin Brown