Prejudice against non-photo camera brands

In article ,
Alan Browne wrote:
John Francis wrote:

My take on in-camera processing is:
Sin #1: image quality is lost when converting to JPG.
Sin #2: scene latitude is lost when converting to JPG.
Sin #3: in camera sharpening is overdone in some cases (including RAW).


Which cameras sharpen a raw image?

I stated that a bit quickly and simply.

From what I gather in other NG discussions, most of them do sharpen at RAW, but
I may be remembering a misperception. I can't cite a source offhand.

It may also, rather than specific sharpening, be artifacts of interpolating
seperate R,G,B sensors into RGB pixels that makes sharpening-like artifacts in
the RAW image.

I think you may have a misunderstanding of just what a RAW image is.
It is a direct measure of the sensor values, prior to any conversion
to RGB pixels. Interpolation artifacts, etc., would be introduced
during processing stages that take place later on in the chain, and
so are not present in the RAW capture.

In general the only camera settings that affect the content of a RAW
image are the effective ISO (maybe including exposure compensation),
white balance (sometimes), and possibly the contrast. Other settings
such as sharpening will be generally be recorded along with the data,
and may very well affect the way the manufacturer-supplied conversion
software behaves, but don't change the recorded pixel values directly.

Adobe have recently announced a new "Digital Negative" file format
(DNG) which attempts to provide a vendor-neutral format for RAW images.
They have a converter to go from the RAW format of many cameras to DNG.
Converting the RAW images from your camera to a well-documented format,
and then seeing just what is in there, is a good way to get an initial
understanding of just what a RAW file contains.

You can also see when and where artifacts are introduced by trying
different conversion software. If you don't have access to a full
version of PhotoShop, PhotoShop Elements 3.0 includes Adobe Camera Raw
(which can read DNG as well as the RAW format of many cameras).

Disclaimer: I am in no way associated with Adobe, except as a customer.
I recently purchased Elements 3.0, and consider it money well spent.

John Francis wrote:

It may also, rather than specific sharpening, be artifacts of interpolating
seperate R,G,B sensors into RGB pixels that makes sharpening-like artifacts in
the RAW image.


I think you may have a misunderstanding of just what a RAW image is.
It is a direct measure of the sensor values, prior to any conversion
to RGB pixels. Interpolation artifacts, etc., would be introduced
during processing stages that take place later on in the chain, and
so are not present in the RAW capture.

In any case, at some point the RAW image may converted to a "lossless" digital
format such as TIFF. In that conversion, regardless of where it takes place,
some artifacts of conversion are introduced. They hardly have a choice but to
be produced as interpolation (of whatever variety) to fill an RGB pixel from
spatially separated pixels must be imperfect.


In general the only camera settings that affect the content of a RAW
image are the effective ISO (maybe including exposure compensation),
white balance (sometimes), and possibly the contrast. Other settings
such as sharpening will be generally be recorded along with the data,
and may very well affect the way the manufacturer-supplied conversion
software behaves, but don't change the recorded pixel values directly.

Please be specific: does sharpening occur in the camera on RAW images?
(automatically or optionally)?



You can also see when and where artifacts are introduced by trying
different conversion software. If you don't have access to a full
version of PhotoShop, PhotoShop Elements 3.0 includes Adobe Camera Raw
(which can read DNG as well as the RAW format of many cameras).

I'll just use the OEM - TIFF converter then continue in PS E 2.

Cheers,
Alan


--
-- r.p.e.35mm user resource: http://www.aliasimages.com/rpe35mmur.htm
-- r.p.d.slr-systems: http://www.aliasimages.com/rpdslrsysur.htm
-- [SI] gallery & rulz: http://www.pbase.com/shootin
-- e-meil: there's no such thing as a FreeLunch.

In article ,
says...
HP sells a wide range of products and makes none of them. They have spun
off their high-end/high-quality/USA-manufactured products into a subsidiary
called Agilent.

This is not true. The Proliant Server line (formerly Compaq, and the Houston
employees still think of it as Compaq, not HP -- watch closely as a current
HP Proliant server boots up, it flashes Compaq for a half second before it is
overwritten with HP text, hehe) is still designed, tested and supported out
of Texas.

Dell makes practically everything in Taiwan or China (does not even do the
board design themselves anymore).

Get used to it, the electronics business is elsewhere. The US will have
to find something else to keep itself busy. Electronics is a commodity
like toaster ovens today.

In article ,
Alan Browne wrote:
John Francis wrote:

It may also, rather than specific sharpening, be artifacts of interpolating
seperate R,G,B sensors into RGB pixels that makes sharpening-like artifacts in
the RAW image.

I think you may have a misunderstanding of just what a RAW image is.
It is a direct measure of the sensor values, prior to any conversion
to RGB pixels. Interpolation artifacts, etc., would be introduced
during processing stages that take place later on in the chain, and
so are not present in the RAW capture.

In any case, at some point the RAW image may converted to a "lossless" digital
format such as TIFF. In that conversion, regardless of where it takes place,
some artifacts of conversion are introduced. They hardly have a choice but to
be produced as interpolation (of whatever variety) to fill an RGB pixel from
spatially separated pixels must be imperfect.

True. But that's going to be the case no matter who does the conversion.
But it's a fair bet that software running on a 3GHz Pentium 4 with 1GB of
RAM, and no real time constraints, can probably do a better job than any
in-camera conversion running on a processor whose main design goal is low
power consumption. Plus, of course, if you delay the conversion to a later
stage you always have the option of trying a different algorithm if you
don't like the effects on any given image.

In general the only camera settings that affect the content of a RAW
image are the effective ISO (maybe including exposure compensation),
white balance (sometimes), and possibly the contrast. Other settings
such as sharpening will be generally be recorded along with the data,
and may very well affect the way the manufacturer-supplied conversion
software behaves, but don't change the recorded pixel values directly.

Please be specific: does sharpening occur in the camera on RAW images?
(automatically or optionally)?

No. It does not. Neither automatically nor optionally.
Sharpening occurs during the conversion from RAW to TIFF/JPG/PSD/...

You can also see when and where artifacts are introduced by trying
different conversion software. If you don't have access to a full
version of PhotoShop, PhotoShop Elements 3.0 includes Adobe Camera Raw
(which can read DNG as well as the RAW format of many cameras).

I'll just use the OEM - TIFF converter then continue in PS E 2.

OK. But that locks you into the Bayer interpolation algorithms chosen
by your OEM, complete with any artifacts introduced during that process.
Some sharpening will be done at this stage of the processing (just how
much may, or may not, depend on the value of the sharpening control set
on the camera at the time of image capture).

John Francis wrote:

True. But that's going to be the case no matter who does the conversion.
But it's a fair bet that software running on a 3GHz Pentium 4 with 1GB of
RAM, and no real time constraints, can probably do a better job than any
in-camera conversion running on a processor whose main design goal is low
power consumption. Plus, of course, if you delay the conversion to a later

Bah! Such simple conversion/interpolation can be done on the fly while storing
to flash with a handful of machine instructions.

stage you always have the option of trying a different algorithm if you
don't like the effects on any given image.

Camera firmware can be upgraded on most DSLR's.

No. It does not. Neither automatically nor optionally.
Sharpening occurs during the conversion from RAW to TIFF/JPG/PSD/...

It doesn't have to happen. There is no need to sharpen an image until the
photog has seen and decides what the USM parameters should be.

You can also see when and where artifacts are introduced by trying
different conversion software. If you don't have access to a full
version of PhotoShop, PhotoShop Elements 3.0 includes Adobe Camera Raw
(which can read DNG as well as the RAW format of many cameras).

I'll just use the OEM - TIFF converter then continue in PS E 2.


OK. But that locks you into the Bayer interpolation algorithms chosen
by your OEM, complete with any artifacts introduced during that process.
Some sharpening will be done at this stage of the processing (just how
much may, or may not, depend on the value of the sharpening control set
on the camera at the time of image capture).

You're confusing me. If no sharpening occurs in camera, then no on the camera
values should be considered.

Cheers,
Alan.


--
-- r.p.e.35mm user resource: http://www.aliasimages.com/rpe35mmur.htm
-- r.p.d.slr-systems: http://www.aliasimages.com/rpdslrsysur.htm
-- [SI] gallery & rulz: http://www.pbase.com/shootin
-- e-meil: there's no such thing as a FreeLunch.

In article ,
Alan Browne wrote:
John Francis wrote:

True. But that's going to be the case no matter who does the conversion.
But it's a fair bet that software running on a 3GHz Pentium 4 with 1GB of
RAM, and no real time constraints, can probably do a better job than any
in-camera conversion running on a processor whose main design goal is low
power consumption. Plus, of course, if you delay the conversion to a later

Bah! Such simple conversion/interpolation can be done on the fly while storing
to flash with a handful of machine instructions.

I really suggest you do a little research into this subject before
dismissing it in quite such a cavalier fashion. If all you want is
some of the most simplistic interpolation, then all it takes is a
few processor cycles. But that also ends up with some of the worst
algorithms. There has been quite a bit of research, and more than a
few thesis papers, done in this field. A good place to start would
be some of the work referenced in the description accompanying dcraw,
but that's only one part in a very complex field. The algorithm
that dcraw uses is more than a plain context-free interpolation, but
it is still a fairly simple algorithm, with limited requirements for
processor resources. There are algorithms with much heavier demands.

stage you always have the option of trying a different algorithm if you
don't like the effects on any given image.

Camera firmware can be upgraded on most DSLR's.

Which just ends up locking you into a different fixed algorithm.
There's no "best" answer that is appropriate in every case

No. It does not. Neither automatically nor optionally.
Sharpening occurs during the conversion from RAW to TIFF/JPG/PSD/...

It doesn't have to happen. There is no need to sharpen an image until the
photog has seen and decides what the USM parameters should be.

You can also see when and where artifacts are introduced by trying
different conversion software. If you don't have access to a full
version of PhotoShop, PhotoShop Elements 3.0 includes Adobe Camera Raw
(which can read DNG as well as the RAW format of many cameras).

I'll just use the OEM - TIFF converter then continue in PS E 2.


OK. But that locks you into the Bayer interpolation algorithms chosen
by your OEM, complete with any artifacts introduced during that process.
Some sharpening will be done at this stage of the processing (just how
much may, or may not, depend on the value of the sharpening control set
on the camera at the time of image capture).

You're confusing me. If no sharpening occurs in camera, then no on the camera
values should be considered.

What's so difficult to understand? You can set a sharpening parameter
on the camera, and the value you set is stored along with the image data.
The later software processing *may* decide to take the value you set as
a starting point to control how much sharpening to apply. Or it may not,
and only sharpen based on values set interactively at that time. Just
which approach is taken depends on how that later stage software is written.

Don't assume all conversion is done interactively, with the photographer
reviewing each image. Sometimes a converter will be run as a batch process
to convert a large number of images. In that case it is sometimes worth
using values selected on-camera (for sharpening, white balance, etc.)
But in no case is any of the sharpening actually performed in-camera.

John Francis wrote:

In article ,
Alan Browne wrote:

John Francis wrote:


True. But that's going to be the case no matter who does the conversion.
But it's a fair bet that software running on a 3GHz Pentium 4 with 1GB of
RAM, and no real time constraints, can probably do a better job than any
in-camera conversion running on a processor whose main design goal is low
power consumption. Plus, of course, if you delay the conversion to a later

Bah! Such simple conversion/interpolation can be done on the fly while storing
to flash with a handful of machine instructions.


I really suggest you do a little research into this subject before
dismissing it in quite such a cavalier fashion. If all you want is
some of the most simplistic interpolation, then all it takes is a
few processor cycles. But that also ends up with some of the worst
algorithms. There has been quite a bit of research, and more than a
few thesis papers, done in this field. A good place to start would
be some of the work referenced in the description accompanying dcraw,
but that's only one part in a very complex field. The algorithm
that dcraw uses is more than a plain context-free interpolation, but
it is still a fairly simple algorithm, with limited requirements for
processor resources. There are algorithms with much heavier demands.

There are always ways to take complex functions and tune for maximum BW in a
constrained case. Been there.

stage you always have the option of trying a different algorithm if you
don't like the effects on any given image.

Camera firmware can be upgraded on most DSLR's.


Which just ends up locking you into a different fixed algorithm.
There's no "best" answer that is appropriate in every case

Or a menu of algorithms. But frankly, beyond a well conceived interpolation of
the R,G,B into RGB, I want to have full control over further sharpenning. (This
attitude, BTW, is the result of having scannned thousands of slides and
negatives. Every image needs USM according to the level of fime detail in the
image.

You're confusing me. If no sharpening occurs in camera, then no on the camera
values should be considered.


What's so difficult to understand? You can set a sharpening parameter
on the camera, and the value you set is stored along with the image data.

Okay, although that sounds a bit silly. Just give me the least processed image
and let me work it over with USM.

The later software processing *may* decide to take the value you set as
a starting point to control how much sharpening to apply. Or it may not,
and only sharpen based on values set interactively at that time. Just
which approach is taken depends on how that later stage software is written.

Don't assume all conversion is done interactively, with the photographer
reviewing each image. Sometimes a converter will be run as a batch process
to convert a large number of images. In that case it is sometimes worth
using values selected on-camera (for sharpening, white balance, etc.)
But in no case is any of the sharpening actually performed in-camera.

Don't assume that I find that acceptable. I'm as lazy as the next guy, but for
detailed images that I want to print large, sharpenning is no less important
than any other aspect of the workflow. I don't want the camera (or the RAW
converter) doing anything to the image that is not reversible. Unless I have
the sharpening algorithm used I can't undo it.

Cheers,
Alan.

--
-- r.p.e.35mm user resource: http://www.aliasimages.com/rpe35mmur.htm
-- r.p.d.slr-systems: http://www.aliasimages.com/rpdslrsysur.htm
-- [SI] gallery & rulz: http://www.pbase.com/shootin
-- e-meil: there's no such thing as a FreeLunch.

In article ,
Alan Browne wrote:
John Francis wrote:

There are always ways to take complex functions and tune for maximum BW in a
constrained case. Been there.

As have I. I've done FFT algorithms on bit-slice processors,
pixel-based algorithms on everything from 68000s to vector
machines, etc. You can, with well-crafted code, get close to
100% utilisation of the hardware. But you can't do better
than that, and some of the reconstruction algorithms simply
need more computational power than is available in-camera.

Or a menu of algorithms. But frankly, beyond a well conceived interpolation of
the R,G,B into RGB, I want to have full control over further sharpenning. (This
attitude, BTW, is the result of having scannned thousands of slides and
negatives. Every image needs USM according to the level of fime detail in the
image.

Bayer reconstruction isn't sharpening; it's rather more complex.
USM is a very simple algorithm, with very little computational load.
It's fairly easy to get control over the sharpening; just tell the
conversion software how much (or how little) sharpening to do. What
you can't do is tell the software which reconstruction algorithm to use.
(That first sentence of yours, containing the phrase "a well-conceived
interpolation of the R,G,B into RGB", is where all the difficulty lies)

Don't assume that I find that acceptable. I'm as lazy as the next guy, but for
detailed images that I want to print large, sharpenning is no less important
than any other aspect of the workflow. I don't want the camera (or the RAW
converter) doing anything to the image that is not reversible. Unless I have
the sharpening algorithm used I can't undo it.

Sharpening isn't reversible, even when you know the algorithm. Nor is
gaussian blur, or almost any of the standard image processing filters.

But I suspect, once we get past the rhetoric, you and I agree on far
more points than we disagree on. We both want the camera to do as
little as possible to the image, and will work by hand to achieve the
best results possible. Just bear in mind that one of the tools we use
along the way, the camera, is designed also to be used by people who
will create prints, if at all, by simply plugging the memory card into
their printer, or a mall photo kiosk. You don't have to use it that
way, and you don't have to use features designed to facilitate that.

Alan Browne wrote:

Or a menu of algorithms. But frankly, beyond a well conceived interpolation
of the R,G,B into RGB, I want to have full control over further sharpenning.

I can't help but wonder, what camera and workflow are you using that doesn't
give you this?

What's so difficult to understand? You can set a sharpening parameter
on the camera, and the value you set is stored along with the image data.

Okay, although that sounds a bit silly. Just give me the least processed
image and let me work it over with USM.

That's, like, exactly what you get. The sharpening setting you choose on
the camera just stores a little setting in the file that tells post-process
software what setting you chose; it doesn't touch the image. The setting
is utterly and completely meaningless, when shooting RAW, in other words.

With Nikon, at least, Adobe Camera Raw can't even extract that setting, so
it might as well not even be there. *None* of the image processing settings
on the camera have any effect whatsoever on the resulting RAW image. You
expose the image and that's it.

I don't want the camera (or the RAW converter) doing anything to the image
that is not reversible.

It doesn't.

--
Jeremy |

"Randy Howard" wrote in message
.. .
In article ,
says...
HP sells a wide range of products and makes none of them. They have spun
off their high-end/high-quality/USA-manufactured products into a
subsidiary
called Agilent.

This is not true. The Proliant Server line (formerly Compaq, and the
Houston
employees still think of it as Compaq, not HP -- watch closely as a
current
HP Proliant server boots up, it flashes Compaq for a half second before it
is
overwritten with HP text, hehe) is still designed, tested and supported
out
of Texas.

And how long do you think that will last? I'd be job hunting if I was
employed there.

Dell makes practically everything in Taiwan or China (does not even do the
board design themselves anymore).

True.

Get used to it, the electronics business is elsewhere. The US will have
to find something else to keep itself busy. Electronics is a commodity
like toaster ovens today.

You are correct but I'll never "get used to it." As we become less and less
self-sufficient, our world position becomes more and more tenuous
(especially so, given our current politcal posture).