EOS 7D and resolution

In article , David J
Taylor
wrote:

"Elliott Roper" wrote in message
...
In article , Elliott Roper
wrote:

In article , Wolfgang
Weisselberg wrote:
snip
RAW gives me 14 bits/channel to play with later. 128 times better than
the first jpg the camera thought of. Why throw away all that
post-processing fun?.
Damn! Would you make that 64 times?

Except that the JPEG contains gamma-corrected data, with a greater
white-black dynamic range than 14-bit raw, albeit at a reduced brightness
accuracy, so you aren't actually comparing like with like.

How can that possibly be so? Where does it get the information from?
And where does it put it? Scrunching it up on a log curve is neat, but
it still does not invent dynamic range from thin air.

I can sure push RAW around in post far more than I can push a jpg. So
the gamma correction does not seem to be delivering even that much.

--
To de-mung my e-mail address:- fsnospam$elliott$$
PGP Fingerprint: 1A96 3CF7 637F 896B C810 E199 7E5C A9E4 8E59 E248

"Elliott Roper" wrote in message
...
[]
How can that possibly be so? Where does it get the information from?
And where does it put it? Scrunching it up on a log curve is neat, but
it still does not invent dynamic range from thin air.

I can sure push RAW around in post far more than I can push a jpg. So
the gamma correction does not seem to be delivering even that much.

Yes, as I said although the dynamic range is greater (it's about
200,000:1) in a pure gamma-corrected JPEG, the accuracy is much less. At
the low end the lowest value 1 is 1/200,000 full scale, and the next value
is 1/50,000 full scale. Similarly at the high end the difference between
254 and 255 is 1%, rather than the 0.4% you might expect.

The gamma correction delivers you data, which is adequate for a human to
view under typical conditions, in a more compact form.

Cheers,
David

In article , David J
Taylor
wrote:

"Elliott Roper" wrote in message
...
[]
How can that possibly be so? Where does it get the information from?
And where does it put it? Scrunching it up on a log curve is neat, but
it still does not invent dynamic range from thin air.

I can sure push RAW around in post far more than I can push a jpg. So
the gamma correction does not seem to be delivering even that much.

Yes, as I said although the dynamic range is greater (it's about
200,000:1) in a pure gamma-corrected JPEG, the accuracy is much less. At
the low end the lowest value 1 is 1/200,000 full scale, and the next value
is 1/50,000 full scale. Similarly at the high end the difference between
254 and 255 is 1%, rather than the 0.4% you might expect.

The gamma correction delivers you data, which is adequate for a human to
view under typical conditions, in a more compact form.

I still don't get it. If the camera is detecting 14 bits per channel,
that's a dynamic range of 2^14 == 1/16384 per channel. Where are the
other 3.61 bits coming from?
I'll accept that if you had 'em in the first place, you could trade
accuracy for range, as you scrunch down to 8 bits, but the camera's
internal jpg processor is only given 14 bits to play with.

As a general rule, the camera's un-informed first guess at a jpg is
pants, and with the compressed pile of bits it hands back, you are
stuck with it.

--
To de-mung my e-mail address:- fsnospam$elliott$$
PGP Fingerprint: 1A96 3CF7 637F 896B C810 E199 7E5C A9E4 8E59 E248

"Elliott Roper" wrote in message
...
[]
I still don't get it. If the camera is detecting 14 bits per channel,
that's a dynamic range of 2^14 == 1/16384 per channel. Where are the
other 3.61 bits coming from?

Because a gamma-corrected JPEG has a non-linear relation between number
and light-level represented. In the 14-bit linear coding, 16383 is the
maximum level, and 8192 half the light level. In the gamma-corrected
JPEG, the mid-value is not 128/255 of the maximum - but (128/255)^2.2 -
about 0.22 of the maximum light level. 1/255 represents about 1/200,000
of the maximum light level.

I'll accept that if you had 'em in the first place, you could trade
accuracy for range, as you scrunch down to 8 bits, but the camera's
internal jpg processor is only given 14 bits to play with.

As a general rule, the camera's un-informed first guess at a jpg is
pants, and with the compressed pile of bits it hands back, you are
stuck with it.

Different cameras make better or worse jobs at converting raw to JPEG. I
am happy with what my cameras do, but appreciate that others may wish to
do the work in a more customised way. I do try and get it "right first
time" in the camera, so that zero or minimum subsequent processing is
required.

Most of my images are intended for display on monitors rather than for
printing, and I find that I can do some post-processing on JPEG images
without seeing too many ill effects. Your needs may well be different.
The image below had quite a boost to the lower light levels to recover the
wood grain, and viewed at this size you are pushed to see any artefacts
resulting from processing the JPEG.

http://www.satsignal.eu/Hols/2009/An...-1230-41-a.jpg

Cheers,
David

In article , David J
Taylor
wrote:

"Elliott Roper" wrote in message
...
[]
I still don't get it. If the camera is detecting 14 bits per channel,
that's a dynamic range of 2^14 == 1/16384 per channel. Where are the
other 3.61 bits coming from?

Because a gamma-corrected JPEG has a non-linear relation between number
and light-level represented. In the 14-bit linear coding, 16383 is the
maximum level, and 8192 half the light level. In the gamma-corrected
JPEG, the mid-value is not 128/255 of the maximum - but (128/255)^2.2 -
about 0.22 of the maximum light level. 1/255 represents about 1/200,000
of the maximum light level.

We are in violent agreement on that part.

I'll accept that if you had 'em in the first place, you could trade
accuracy for range, as you scrunch down to 8 bits, but the camera's
internal jpg processor is only given 14 bits to play with.
So the jpg processor in the camera is making wild guesses at mapping
the 14 bits of real level all over its pretend range of 17.6 bits as it
assigns its guess to one of 256 values? That's log(2) of 200,000 for
everyone else following this.
*That's* what you meant in your first post. I get it now.
As a general rule, the camera's un-informed first guess at a jpg is
pants, and with the compressed pile of bits it hands back, you are
stuck with it.

Different cameras make better or worse jobs at converting raw to JPEG. I
am happy with what my cameras do, but appreciate that others may wish to
do the work in a more customised way. I do try and get it "right first
time" in the camera, so that zero or minimum subsequent processing is
required.
Yep. My 5Dii sometimes does a great job of jpeg, but usually does not.
For me, the cost in time and lost opportunity favours shooting RAW and
fixing it later. Most of the time I don't bother with bracketing
exposure, or messing with white balance, or dialling in some offset or
other, because the subject won't wait. RAW gives me everything the
camera can give at a particular exposure.
Most of my images are intended for display on monitors rather than for
printing, and I find that I can do some post-processing on JPEG images
without seeing too many ill effects. Your needs may well be different.
The image below had quite a boost to the lower light levels to recover the
wood grain, and viewed at this size you are pushed to see any artefacts
resulting from processing the JPEG.

http://www.satsignal.eu/Hols/2009/An...-1230-41-a.jpg

I love the light on the Ovaltine tin fragment compared to the twinkles
on the badly rusted other stuff. If I were shooting that (and if I had
the art to spot the opportunity) it would have been one for exposure
bracketing even with RAW. I revel in the chance to fix shadows and
highlights, play with levels, gently tweak the definition before
finishing with a jpeg export. Starting with RAW gives me so much more
sand in my sandpit at playtime.

Thanks for your patience with me on this. I'm off to do some jpeg+RAW
shooting to compare your methods and mine. In true Antarctic fashion "I
may be gone some time"

--
To de-mung my e-mail address:- fsnospam$elliott$$
PGP Fingerprint: 1A96 3CF7 637F 896B C810 E199 7E5C A9E4 8E59 E248

"Elliott Roper" wrote in message
...
[]
So the jpg processor in the camera is making wild guesses at mapping
the 14 bits of real level all over its pretend range of 17.6 bits as it
assigns its guess to one of 256 values? That's log(2) of 200,000 for
everyone else following this.
*That's* what you meant in your first post. I get it now.

Well, let's hope it's not a wild guess, but a well-defined lookup! Of
course the raw has a greater precision in defining the values, allowing
much more post processing, and it will have more headroom as well.

Yep. My 5Dii sometimes does a great job of jpeg, but usually does not.
For me, the cost in time and lost opportunity favours shooting RAW and
fixing it later. Most of the time I don't bother with bracketing
exposure, or messing with white balance, or dialling in some offset or
other, because the subject won't wait. RAW gives me everything the
camera can give at a particular exposure.

Yes, I can see how that would work for you - saving time during taking.



http://www.satsignal.eu/Hols/2009/An...-1230-41-a.jpg

I love the light on the Ovaltine tin fragment compared to the twinkles
on the badly rusted other stuff. If I were shooting that (and if I had
the art to spot the opportunity) it would have been one for exposure
bracketing even with RAW. I revel in the chance to fix shadows and
highlights, play with levels, gently tweak the definition before
finishing with a jpeg export. Starting with RAW gives me so much more
sand in my sandpit at playtime.

Thanks for your patience with me on this. I'm off to do some jpeg+RAW
shooting to compare your methods and mine. In true Antarctic fashion "I
may be gone some time"

It sounds as if we both enjoy what we're doing, and that's the main thing
for me as an amateur photographer. I like to understand as much as I can,
coming from both a TV and a digital signal processing background.

Cheers,
David

Elliott Roper wrote:
In article , Wolfgang
Elliott Roper wrote:
Wolfgang Weisselberg wrote:

Heh! You are just winding me up for sport aren't you? But you do it
with style, so I'll play.

How many cards per camera do you use and how many did you use at
most at any time?

1 and a little 1

Hmmm ... while most cameras only allow one card inserted, I find
that I use quite more than one card per DSLR ... and I'd be using
a much higher number of cards if I didn't "recycle" them through
the image tank.

Googling seems to show $400 for a shutter replacement. USD0.003 per
click with infinite body life. You win!

Winning never was so cheap (or is tha devalued?) than with digital.

Seriously, the most likely risk of data loss is
down to old butterfingers me.

Of course.

I'll drop a drive,

That's why I don't belive in backup hard disks --- especially
movable ones --- and use tapes for that. (Or internet backup.)

re-initialise the
wrong one, delete the wrong project 1000 times more frequently than the
burglars and the arsonists come to visit.

Of course.
Still, you don't take out insurance for the common, low cost
accidents in life but for the big ones. Even if you hope they'll
never happen.

Do you have an extra fast computer? One that is overpowered

No. I use a very ordinary 2006 Mac Pro with Aperture and the standard
[...]
Poor little thing works quite hard
rendering and colour correcting stuff while I sleep.

I see. Fine, if you don't need interactive tuning with your
rendering and colour correcting --- or if a turn-around time of
a day is OK with you. Unfortunately, I am much more demanding
than that --- if I change a slider, I don't want to wait till
tomorrow to see the results.

Overnight final colour correction for an hour of HD movie I can live
with.

But you still need pretty instant response to your preview ---
be it a single frame or a low resolution version --- so you can
judge the colour correction.

I was answering your point about economically over-configuring a
computer that is infrequently used for still image processing. I too
expect a slider to slide.

And the effect behind that slider to happen, too. Imagine sliding
the "fill-in lights" slider and waiting for a minute till the
effect becomes visible!

So you say you would inspect a 360x240 pixel image as painstakingly
as a gigapixel panorama? I disbelieve!

You are welcome to. I think my 'inspect' and yours differ. Since I did
not define mine I can hardly complain. Mine in that context was "Do I
bother with this shot?"

Which is a complete different thing from "Does this (hopefully
final) version of the shot have any nasty bad pixels or other
visually detracting 'features' one might see when inspecting
e.g. the print in detail?".

RAW decodes are about 1 second,

You meant to say "The batch RAW decoding I use with standard
parameters on my machine ..." actually, your camera is faster
than that, and gives you a complete JPEG to enjoy, not just
another uncooked format that's blown up by having all 3 channels
in the TIFF.

Now you are /really/ in it for the sport.

Meee??

I did not mean to say
anything of the sort.
RAW gives me 14 bits/channel to play with later.

How many of these 14 bits are signal? I remember reading
that even current cameras that deliver 14 bit don't deliver
more than 12 bits of real signal ...

128 times better than
the first jpg the camera thought of. Why throw away all that
post-processing fun?.

.... and of course you can do a 16 or 20 or 32 bit-per-channel
postprocessing even when the source signal has much less than
that amount bits. Simply fill up the insiginifcant rest with
noise. :-)

So you review 1000 RAWs in about an hour or three? 4 - 10
seconds for each shot? Hmhm. Can be done, but it's hard work
--- and works only for a first selection. Try "I've got a couple
thousand shots, unsorted --- is this shot the best one of person X
(one of the 100 people attending)?".

Sure, I was talking first selection. What do you mean 1000 shots
unsorted? That's what computers are for. Face detect,

So your computer does detect a *specific* face out of hundreds
possible in any orientation (e.g. full frontal to side profile),
even half-hidden by musical instruments?

Otherwise --- yes, the shots nearly all have faces in them.
Congrats, you sorted everything.

Stacks - auto or otherwise,

None yet.

rating,

None yet.

EXIF search,

Have fun, you'll not find much ...

keywords

None yet.

-- 1000 shots don't stay unsorted for long.

I'd like the shots sorted by the main person on the image,
subsorted by the visual impact.

My original point was that a combination of an obscene number of
megapixels, great prime lenses and a decent lump of image processing
software makes a very usable and versatile system.

Depends on what your needs are. If your needs are along the lines
of quickly changing focal lengths, 4x6" prints out of the camera,
low weight, small packsize ... you loose.

Definitely not my needs. 4x6 inch prints straight from the camera? Moi?

You're not selling photographs from your mobile printer??

Anyway it's an ... interesting ... attitude to use "great
prime lenses" and then throw away their advantage over any
*bad* zoom lens by cropping "like crazy".

I don't buy the monkish attitude that every shot has to be composed
carefully before clicking.

Why should composing properly be "monkish"?

Look, I never said "composing properly" I said "composing carefully
before clicking".

Ok, but what's "monkish" about composing properly before
clicking?

There is nothing wrong with composing in post, as
long as you have enough (ahem) RAW material to work with.

Of course there's not. But it's a step more, and if you can
avoid another step of work because you become more proficient
in using your tools, it's a reason to be glad.

I'm beginning
to regret using "cropping like crazy" in this discussion. I should have
been less 'entertaining' with "lightly cropping to improve composition
without grossly compromising image quality" by taking advantage of good
glass and good sensors and an adequate post-processing computer.

Lightly cropping to me is maybe 10-15% of each border, which,
worst case, removes 50% of the pixels already!

Cropping like crazy starts at taking a 2:3 portrait shot, cropping
it into 2:3 landscape orientation, and then removing lots at the
borders, keeping less than 20-25% of the pixels.

And in the latter case, unless you *cannot* get physically
closer and *cannot* use a longer lens, is a sure recipe of using
extraorinary glass to get results ordinary glass can surpass.

It is still art and skill.

Painting the scene of an accident for the insurance company
instead of using the available and sufficient camera image
also creates art and demands skill. It's just that the
faster, better, cheaper way also produces superior results
for the given needs.

Why don't you just use a fisheye and fixed ISO 100, 1/100s,
f/16 and do everything in postprocessing? That's the logical
solution to not composing, metering etc. at all! That's
surely how great artists work!

That's down there with the "great artists only use Box
Brownies" myth.

Of course, lomography doesn't do composing either, carefully
or not.

Oh come on Wolfgang! I'd wear a 360° continuously recording "lifecam"
if I could push the tech that far.

You can. It's a question of chequebook limits, not one of
technical limits.

Your "logical solution" is a reducto ad ridiculum and you well know
it.

Of course it is. You now get to tell me at what exact point your
art form stops and my ridiculous solution begins ...

I had to Google for "lumography". I confess I first conflated it with
"lupography" where someone whose name begins with "wolf" does a "don't
think, just post".

Nope, lupography is drawing with a wolf. Be careful, it may
bite, and then it's lycantropy.

For someone whose first language is probably not english you are
pretty good.

I really need to work at my accent, then!

-Wolfgang

In article , Wolfgang
Weisselberg wrote:

Elliott Roper wrote:
snip
Of course it is. You now get to tell me at what exact point your
art form stops and my ridiculous solution begins ...
That is far too hard. I guess the serious answer is how horrible each
result will be compared to its intended purpose, but that is far too
sensible and boring.
I had to Google for "lumography". I confess I first conflated it with
"lupography" where someone whose name begins with "wolf" does a "don't
think, just post".

Nope, lupography is drawing with a wolf. Be careful, it may
bite, and then it's lycantropy.


For someone whose first language is probably not english you are
pretty good.

I really need to work at my accent, then!
You do! It's lycanthropy.

Maybe we can call pixel-peepers "Loupe-Garou"?

--
To de-mung my e-mail address:- fsnospam$elliott$$
PGP Fingerprint: 1A96 3CF7 637F 896B C810 E199 7E5C A9E4 8E59 E248

Elliott Roper wrote:
In article , Wolfgang
Weisselberg wrote:

Elliott Roper wrote:
snip
Of course it is. You now get to tell me at what exact point your
art form stops and my ridiculous solution begins ...

That is far too hard.

So you are already on a slippery slope to ridiculousness with
your cropping?

I guess the serious answer is how horrible each
result will be compared to its intended purpose, but that is far too
sensible and boring.

Indeed.

For someone whose first language is probably not english you are
pretty good.

I really need to work at my accent, then!
You do! It's lycanthropy.

*Hoooooowl!*

Maybe we can call pixel-peepers "Loupe-Garou"?

Loupe-Garotte?

-Wolfgang

In article , Wolfgang
Weisselberg wrote:

Elliott Roper wrote:
snip
*Hoooooowl!*

Maybe we can call pixel-peepers "Loupe-Garou"?

Loupe-Garotte?

You deserve the final word with one as good one as that.

--
To de-mung my e-mail address:- fsnospam$elliott$$
PGP Fingerprint: 1A96 3CF7 637F 896B C810 E199 7E5C A9E4 8E59 E248