different max jpg size

Starting with the same tif file, and converting
to the maximum quality jpg image, I get these
approximate sizes:

Capture NX 2: 10,480K
Photoshop CS3: 10,394K
ImageMagick convert: 17,102K

So what do you think of this? I am guessing
that the 17M file has a higher quality image.
Do you agree?

Thanks for your thoughts.
Mike.

On 10-05-02 17:31 , Michael D. Berger wrote:
Starting with the same tif file, and converting
to the maximum quality jpg image, I get these
approximate sizes:

Capture NX 2: 10,480K
Photoshop CS3: 10,394K
ImageMagick convert: 17,102K

So what do you think of this? I am guessing
that the 17M file has a higher quality image.
Do you agree?

First off, what NX 2, CS3 and IM consider to be 7/10 or 70/100 quality
may not be at all the same thing.

WRT to IM, it could be sub-optimal JPG encoding where the image may look
exactly the same as from NX 2 or CS2 but do so with more data. IOW
where it had the opportunity to compress w/o loss, it didn't.


--
gmail originated posts are filtered due to spam.

On 3/05/2010 9:34 a.m., Alan Browne wrote:
On 10-05-02 17:31 , Michael D. Berger wrote:
Starting with the same tif file, and converting
to the maximum quality jpg image, I get these
approximate sizes:

Capture NX 2: 10,480K
Photoshop CS3: 10,394K
ImageMagick convert: 17,102K

So what do you think of this? I am guessing
that the 17M file has a higher quality image.
Do you agree?

First off, what NX 2, CS3 and IM consider to be 7/10 or 70/100 quality
may not be at all the same thing.

WRT to IM, it could be sub-optimal JPG encoding where the image may look
exactly the same as from NX 2 or CS2 but do so with more data. IOW where
it had the opportunity to compress w/o loss, it didn't.


You might be able to get some quantitative analysis by using something
like the "colorcube anaylsis" function of GIMP, and counting number of
unique colours present in the jpegs. Well before visible jpeg artefacts
show, there's usually a large decrease in the number of unique colours
saved, a bad starting point for future editing if posterisation is to be
avoided.
As well as nominal "% quality" setting, there are several parameters for
jpeg compression affecting quality of the saved image, subsampling
method etc.
One potential gotcha if wishing to save images with no losses is
colourspace conversion where you'll find that conversion say from aRGB
to sRGB or vice-versa using PS's conversion engine loses a lot of colour
data. It's one good reason IMO to rely on the original raw format for
archive. I don't use DNG - perhaps that's a better non-proprietary
format than tiff.

"Michael D. Berger" wrote in news:vamDn.113968
:

Starting with the same tif file, and converting
to the maximum quality jpg image, I get these
approximate sizes:

Capture NX 2: 10,480K
Photoshop CS3: 10,394K
ImageMagick convert: 17,102K

So what do you think of this? I am guessing
that the 17M file has a higher quality image.
Do you agree?

No. If it has a higher quality image, you should be able to prove it.
There's no need to guess.

On 10-05-02 18:18 , Me wrote:
On 3/05/2010 9:34 a.m., Alan Browne wrote:
On 10-05-02 17:31 , Michael D. Berger wrote:
Starting with the same tif file, and converting
to the maximum quality jpg image, I get these
approximate sizes:

Capture NX 2: 10,480K
Photoshop CS3: 10,394K
ImageMagick convert: 17,102K

So what do you think of this? I am guessing
that the 17M file has a higher quality image.
Do you agree?

First off, what NX 2, CS3 and IM consider to be 7/10 or 70/100 quality
may not be at all the same thing.

WRT to IM, it could be sub-optimal JPG encoding where the image may look
exactly the same as from NX 2 or CS2 but do so with more data. IOW where
it had the opportunity to compress w/o loss, it didn't.


You might be able to get some quantitative analysis by using something
like the "colorcube anaylsis" function of GIMP, and counting number of
unique colours present in the jpegs. Well before visible jpeg artefacts
show, there's usually a large decrease in the number of unique colours
saved, a bad starting point for future editing if posterisation is to be
avoided.

I would be delighted to read someone else's practical research into the
matter .... ;-)

As well as nominal "% quality" setting, there are several parameters for
jpeg compression affecting quality of the saved image, subsampling
method etc.
One potential gotcha if wishing to save images with no losses is
colourspace conversion where you'll find that conversion say from aRGB
to sRGB or vice-versa using PS's conversion engine loses a lot of colour
data. It's one good reason IMO to rely on the original raw format for
archive. I don't use DNG - perhaps that's a better non-proprietary
format than tiff.

I use compressed DNG which is lossless (per Adobe). Reduces my raw's
from about 33 MB to 15 - 18 MB.

DNG is a wrapper around a TIFF/EP (not 'old' TIFF) version of the camera
raw.

I take lossless to exclude any notion (in DNG) of colorspace, which
makes sense since the image file in the DNG represents the camera raw -
no colorspace at all. (Although the camera colorspace and other setting
data are also contained and available to the raw converter as a 'camera
shot starting point' - the image data is as it was read from the sensor
[as far as that is true camera raw to camera raw]).

DNG is Adobe proprietary though it is open / royalty free. It has been
submitted as an ISO standard (as part of TIFF/EP per the Wiki article),
which would make it completely non-proprietary.

Same article seems unsure about the ISO outcome.

--
gmail originated posts are filtered due to spam.

On 4/05/2010 8:11 a.m., Alan Browne wrote:
On 10-05-02 18:18 , Me wrote:
On 3/05/2010 9:34 a.m., Alan Browne wrote:
On 10-05-02 17:31 , Michael D. Berger wrote:
Starting with the same tif file, and converting
to the maximum quality jpg image, I get these
approximate sizes:

Capture NX 2: 10,480K
Photoshop CS3: 10,394K
ImageMagick convert: 17,102K

So what do you think of this? I am guessing
that the 17M file has a higher quality image.
Do you agree?

First off, what NX 2, CS3 and IM consider to be 7/10 or 70/100 quality
may not be at all the same thing.

WRT to IM, it could be sub-optimal JPG encoding where the image may look
exactly the same as from NX 2 or CS2 but do so with more data. IOW where
it had the opportunity to compress w/o loss, it didn't.


You might be able to get some quantitative analysis by using something
like the "colorcube anaylsis" function of GIMP, and counting number of
unique colours present in the jpegs. Well before visible jpeg artefacts
show, there's usually a large decrease in the number of unique colours
saved, a bad starting point for future editing if posterisation is to be
avoided.

I would be delighted to read someone else's practical research into the
matter .... ;-)

Relatively easy to test yourself. Save an image file (from lossless
format) as jpeg at various compressions, open in Gimp, then run
colorcube analysis plugin. It counts colours and produces a graph.
Problem is that then it gets subjective. A lot of colour data can be
discarded (IMO) before there's noticeable image degradation.
But there are lots of annoyances - one that gets to me for web use is
that blocking/posterisation of gradients (and noise) is usually much
less visible on good quality 8 bit LCD panels or CRTs (or prints for
that matter) than on the 6 bit LCDs almost everyone else uses, but the
quality of dithering used on 6 bit panels also varies widely, so there's
no easy lowest common denominator reference point. I should use my
laptop screen for compressing images for the web.

As well as nominal "% quality" setting, there are several parameters for
jpeg compression affecting quality of the saved image, subsampling
method etc.
One potential gotcha if wishing to save images with no losses is
colourspace conversion where you'll find that conversion say from aRGB
to sRGB or vice-versa using PS's conversion engine loses a lot of colour
data. It's one good reason IMO to rely on the original raw format for
archive. I don't use DNG - perhaps that's a better non-proprietary
format than tiff.

I use compressed DNG which is lossless (per Adobe). Reduces my raw's
from about 33 MB to 15 - 18 MB.

DNG is a wrapper around a TIFF/EP (not 'old' TIFF) version of the camera
raw.

I take lossless to exclude any notion (in DNG) of colorspace, which
makes sense since the image file in the DNG represents the camera raw -
no colorspace at all. (Although the camera colorspace and other setting
data are also contained and available to the raw converter as a 'camera
shot starting point' - the image data is as it was read from the sensor
[as far as that is true camera raw to camera raw]).

DNG is Adobe proprietary though it is open / royalty free. It has been
submitted as an ISO standard (as part of TIFF/EP per the Wiki article),
which would make it completely non-proprietary.

Same article seems unsure about the ISO outcome.

Thanks - I didn't know that there was a lossless compressed dng format.
As a Nikon user, there's less incentive to go the dng way for file
size reason, as compressed lossy or lossless raw formats are possible
ex-camera, depending on model.

On 10-05-03 17:38 , Me wrote:
On 4/05/2010 8:11 a.m., Alan Browne wrote:
On 10-05-02 18:18 , Me wrote:
On 3/05/2010 9:34 a.m., Alan Browne wrote:
On 10-05-02 17:31 , Michael D. Berger wrote:
Starting with the same tif file, and converting
to the maximum quality jpg image, I get these
approximate sizes:

Capture NX 2: 10,480K
Photoshop CS3: 10,394K
ImageMagick convert: 17,102K

So what do you think of this? I am guessing
that the 17M file has a higher quality image.
Do you agree?

First off, what NX 2, CS3 and IM consider to be 7/10 or 70/100 quality
may not be at all the same thing.

WRT to IM, it could be sub-optimal JPG encoding where the image may
look
exactly the same as from NX 2 or CS2 but do so with more data. IOW
where
it had the opportunity to compress w/o loss, it didn't.


You might be able to get some quantitative analysis by using something
like the "colorcube anaylsis" function of GIMP, and counting number of
unique colours present in the jpegs. Well before visible jpeg artefacts
show, there's usually a large decrease in the number of unique colours
saved, a bad starting point for future editing if posterisation is to be
avoided.

I would be delighted to read someone else's practical research into the
matter .... ;-)

Relatively easy to test yourself.

Like I said above ...

Save an image file (from lossless
format) as jpeg at various compressions, open in Gimp, then run
colorcube analysis plugin. It counts colours and produces a graph.
Problem is that then it gets subjective. A lot of colour data can be
discarded (IMO) before there's noticeable image degradation.
But there are lots of annoyances - one that gets to me for web use is
that blocking/posterisation of gradients (and noise) is usually much
less visible on good quality 8 bit LCD panels or CRTs (or prints for
that matter) than on the 6 bit LCDs almost everyone else uses, but the
quality of dithering used on 6 bit panels also varies widely, so there's
no easy lowest common denominator reference point. I should use my
laptop screen for compressing images for the web.

As well as nominal "% quality" setting, there are several parameters for
jpeg compression affecting quality of the saved image, subsampling
method etc.
One potential gotcha if wishing to save images with no losses is
colourspace conversion where you'll find that conversion say from aRGB
to sRGB or vice-versa using PS's conversion engine loses a lot of colour
data. It's one good reason IMO to rely on the original raw format for
archive. I don't use DNG - perhaps that's a better non-proprietary
format than tiff.

I use compressed DNG which is lossless (per Adobe). Reduces my raw's
from about 33 MB to 15 - 18 MB.

DNG is a wrapper around a TIFF/EP (not 'old' TIFF) version of the camera
raw.

I take lossless to exclude any notion (in DNG) of colorspace, which
makes sense since the image file in the DNG represents the camera raw -
no colorspace at all. (Although the camera colorspace and other setting
data are also contained and available to the raw converter as a 'camera
shot starting point' - the image data is as it was read from the sensor
[as far as that is true camera raw to camera raw]).

DNG is Adobe proprietary though it is open / royalty free. It has been
submitted as an ISO standard (as part of TIFF/EP per the Wiki article),
which would make it completely non-proprietary.

Same article seems unsure about the ISO outcome.

Thanks - I didn't know that there was a lossless compressed dng format.
As a Nikon user, there's less incentive to go the dng way for file size
reason, as compressed lossy or lossless raw formats are possible
ex-camera, depending on model.

On my Maxxum 7D I would get a reasonably compressed lossless raw from
the camera (about 25-30%), but on the a900 it is only about 10%.

I surmise that given the huge file size of the a900 raw, they
de-optimized the lossless compression in order to speed up storage/save
the battery. This may explain why compressing the a900 raw to DNG
results in such a large space saving.

With NEF there is a savings going to DNG, just not as much. But then,
given the above it may be just a case of Nikon doing more lossless
compression in-camera.


--
gmail originated posts are filtered due to spam.

On 02 May 2010 21:31:07 GMT, "Michael D. Berger"
wrote:

Starting with the same tif file, and converting
to the maximum quality jpg image, I get these
approximate sizes:

Capture NX 2: 10,480K
Photoshop CS3: 10,394K
ImageMagick convert: 17,102K

So what do you think of this? I am guessing
that the 17M file has a higher quality image.
Do you agree?

Thanks for your thoughts.
Mike.

To find out what is happening try using this:

http://www.impulseadventure.com/photo/jpeg-snoop.html

"JPEGsnoop reports a huge amount of information, including: quantization
table matrix (chrominance and luminance), chroma subsampling, estimates
JPEG Quality setting, JPEG resolution settings, Huffman tables, EXIF
metadata, Makernotes, RGB histograms, etc. Most of the JPEG JFIF markers
are reported. In addition, you can enable a full huffman VLC decode, which
will help those who are learning about JPEG compression and those who are
writing a JPEG decoder."

There was another DOS (command prompt) program called "JPEG Analyzer"
(filename of ja001.zip) that used to be floating around, but it seems to
have disappeared from the net. It would give you the quantization and
huffman tables for any analyzed JPG file.

I found the above just now when searching for the old DOS command-line
program.

On 10-05-03 17:40 , John A. wrote:
On Mon, 03 May 2010 16:11:19 -0400, Alan Browne
wrote:

On 10-05-02 18:18 , Me wrote:
On 3/05/2010 9:34 a.m., Alan Browne wrote:
On 10-05-02 17:31 , Michael D. Berger wrote:
Starting with the same tif file, and converting
to the maximum quality jpg image, I get these
approximate sizes:

Capture NX 2: 10,480K
Photoshop CS3: 10,394K
ImageMagick convert: 17,102K

So what do you think of this? I am guessing
that the 17M file has a higher quality image.
Do you agree?

First off, what NX 2, CS3 and IM consider to be 7/10 or 70/100 quality
may not be at all the same thing.

WRT to IM, it could be sub-optimal JPG encoding where the image may look
exactly the same as from NX 2 or CS2 but do so with more data. IOW where
it had the opportunity to compress w/o loss, it didn't.


You might be able to get some quantitative analysis by using something
like the "colorcube anaylsis" function of GIMP, and counting number of
unique colours present in the jpegs. Well before visible jpeg artefacts
show, there's usually a large decrease in the number of unique colours
saved, a bad starting point for future editing if posterisation is to be
avoided.

I would be delighted to read someone else's practical research into the
matter .... ;-)

As well as nominal "% quality" setting, there are several parameters for
jpeg compression affecting quality of the saved image, subsampling
method etc.
One potential gotcha if wishing to save images with no losses is
colourspace conversion where you'll find that conversion say from aRGB
to sRGB or vice-versa using PS's conversion engine loses a lot of colour
data. It's one good reason IMO to rely on the original raw format for
archive. I don't use DNG - perhaps that's a better non-proprietary
format than tiff.

I use compressed DNG which is lossless (per Adobe). Reduces my raw's
from about 33 MB to 15 - 18 MB.

DNG is a wrapper around a TIFF/EP (not 'old' TIFF) version of the camera
raw.

I take lossless to exclude any notion (in DNG) of colorspace, which
makes sense since the image file in the DNG represents the camera raw -
no colorspace at all. (Although the camera colorspace and other setting
data are also contained and available to the raw converter as a 'camera
shot starting point' - the image data is as it was read from the sensor
[as far as that is true camera raw to camera raw]).

I would think it would have to have some representation of the camera
sensor's native color space. Otherwise, how would you white balance?
Particularly in an open format that's supposed to support any camera's
output.

As I said above, such information is included in the raw data image
_file_, but the raw image data is raw numbers from the sensor untouched.

(However - in later Sony cameras even the raw data may have h/w based
noise filtering included, not sure what alterations other co's may be
doing to the raw prior to storage).


--
gmail originated posts are filtered due to spam.

On 10-05-03 22:14 , John A. wrote:
On Mon, 03 May 2010 18:13:58 -0400, Alan Browne
wrote:

On 10-05-03 17:40 , John A. wrote:
On Mon, 03 May 2010 16:11:19 -0400, Alan Browne
wrote:

On 10-05-02 18:18 , Me wrote:
On 3/05/2010 9:34 a.m., Alan Browne wrote:
On 10-05-02 17:31 , Michael D. Berger wrote:
Starting with the same tif file, and converting
to the maximum quality jpg image, I get these
approximate sizes:

Capture NX 2: 10,480K
Photoshop CS3: 10,394K
ImageMagick convert: 17,102K

So what do you think of this? I am guessing
that the 17M file has a higher quality image.
Do you agree?

First off, what NX 2, CS3 and IM consider to be 7/10 or 70/100 quality
may not be at all the same thing.

WRT to IM, it could be sub-optimal JPG encoding where the image may look
exactly the same as from NX 2 or CS2 but do so with more data. IOW where
it had the opportunity to compress w/o loss, it didn't.


You might be able to get some quantitative analysis by using something
like the "colorcube anaylsis" function of GIMP, and counting number of
unique colours present in the jpegs. Well before visible jpeg artefacts
show, there's usually a large decrease in the number of unique colours
saved, a bad starting point for future editing if posterisation is to be
avoided.

I would be delighted to read someone else's practical research into the
matter .... ;-)

As well as nominal "% quality" setting, there are several parameters for
jpeg compression affecting quality of the saved image, subsampling
method etc.
One potential gotcha if wishing to save images with no losses is
colourspace conversion where you'll find that conversion say from aRGB
to sRGB or vice-versa using PS's conversion engine loses a lot of colour
data. It's one good reason IMO to rely on the original raw format for
archive. I don't use DNG - perhaps that's a better non-proprietary
format than tiff.

I use compressed DNG which is lossless (per Adobe). Reduces my raw's
from about 33 MB to 15 - 18 MB.

DNG is a wrapper around a TIFF/EP (not 'old' TIFF) version of the camera
raw.

I take lossless to exclude any notion (in DNG) of colorspace, which
makes sense since the image file in the DNG represents the camera raw -
no colorspace at all. (Although the camera colorspace and other setting
data are also contained and available to the raw converter as a 'camera
shot starting point' - the image data is as it was read from the sensor
[as far as that is true camera raw to camera raw]).

I would think it would have to have some representation of the camera
sensor's native color space. Otherwise, how would you white balance?
Particularly in an open format that's supposed to support any camera's
output.

As I said above, such information is included in the raw data image
_file_, but the raw image data is raw numbers from the sensor untouched.

(However - in later Sony cameras even the raw data may have h/w based
noise filtering included, not sure what alterations other co's may be
doing to the raw prior to storage).

Ah, so you did. Sorry, I was a bit dehydrated at the time and must
have misparsed "camera colorspace" as "camera white balance", or maybe
as meaning a default working color space selected in-camera. (Mine
lets you select sRGB or aRGB, but that's really only meaningful for
camera-produced jpegs. The raw "developer" program does let you
optionally go along with whichever was selected in the camera for a
particular shot.)

Yep, it's usually, in fact, the default, though with the way Sony embed
some parameters (notably colour temp from the WB) these don't come out
often in ACR as expected. Even if I tell the camera "this is flash" or
flash temp is 5600K with neutral tint - in ACR the temp may be different
by 100 - 200K and the tint off by a few notches (usually to magenta).

Not sure how Nikon/Canon raws and ACR behave.



--
gmail originated posts are filtered due to spam.