Highlight recovery

[jd_davis]

of whether or not you can pull more info back into the RAW image, but what I would have liked to see is whether or not you were able to color correct the second image after getting much of the details back.

Your situation in these photos is extreme obviously. If we can't get a photo to look normal after doing all the processing to retrieve the details, then it's a total waste of time.

I for one have been doing some testing with the metering on the E510. For now it looks like staying on the negative side of 0.0 a little (on a normal day and normal light) seems to work better. The camera is sensative with exposure in jpg, but should be easier to work with photos shot in RAW. (I haven't tested much in RAW yet but I will soon.)

Hopefully we will never need so much dynamic range that we would need to fix photos that look this bad..!! : )) We need color information too.

--
jd
---

 

[rovingtim]

I think camera 2 is full frame. But, of course, we must wait for the MASTER to confirm.

Off topic, for anyone interested, professional photographer has an interesting comparison of the D3 vs the 1Ds MIII

 

[Big Ga]

Hi Gareth

Interesting that both cameras show banding.

I think ALL cameras show banding. it just depends on when it shows.

The fact that I managed to get camera2 to indeed show it just demonstrates how severe the pulling back in PP was - far more than one would ever have to do - hence why I said I should try and shoot something sensible like some black and white clothing next to one another.

When I did my 'studio tests' over Xmas, I really screwed up and forgot to deliberately overexpose every camera by the same amount so I had a more controlled scenario. I was hoping something in the scene was blown, but hardly anything was. Oh well.... live and learn.

 

[Big Ga]

Hopefully we will never need so much dynamic range that we would need
to fix photos that look this bad..!! : ))

Oh trust me .... I've been there and done it!

bang - you grab a camera from the bag, just manage to get it to your eye and press the trigger before the moment is lost and ..... you forgot you left it in manual mode and metered for inside the building you left ...

or bang - the flash misfired .... but that was the shot where the expression on the faces was the best one ...

We need color information
too.

Hey ... that's what B&W is for ....

 

[rovingtim]

I think ALL cameras show banding. it just depends on when it shows.

You mean all CMOS cameras, don't you? I just saw one of my old E1 3200 ISO shots and was thinking, Hey, that's not too bad!

And there was NO BANDING!

 

[Big Ga]

Off topic, for anyone interested, professional photographer has an
interesting comparison of the D3 vs the 1Ds MIII

Yea ... some might say its a bit strange comparing two cameras aimed at two very different markets. but hey ... its two 'flagship cameras' up against one another. I found it very interesting (even though I thought they probably screwed Canon as they MUST have used an iffy copy of the 28-70 2.8L lens)

Can you imagine the uproar here if someone was to pit the Olympus flagship directly up against the Canik ones ;-)

Is anyone complaining on the other forums? I haven't looked.

G.

 

[rovingtim]

Did you notice their response to this criticism? Check the letter at the beginning of the mag ...

 

[Big Ga]

I think ALL cameras show banding. it just depends on when it shows.

You mean all CMOS cameras, don't you? I just saw one of my old E1
3200 ISO shots and was thinking, Hey, that's not too bad!

And there was NO BANDING!

Hmm.. good point. I don't think I've ever pushed an E1/E300 that far though

Probably worth giving it a go. I shudder to think what the real equivalent exposure compensation was in my test ... may have been something daft like 6 stops or more. Its hard to tell. Hence why I'll try (but no promises) to do some more tomorrow.

G.

 

[rovingtim]

Well, I agree that the old CCD's would produce nothing more than a mass of colour mush if you pushed them that far. But they wouldn't band. So there!

 

[Big Ga]

Did you notice their response to this criticism? Check the letter at
the beginning of the mag ...

Ok, had the mag here next to me. I see it. Interesting. So res doesn't necessarily matter eh ;-)

 

[rovingtim]

I get great dynamic range with my 1 byte camera.

Well, okay, I actually agree with them. The high ISO resolution tests I thought were very interesting. It's the overall IQ that matters and resolution is only one factor. 10MP must be good enough for 95% of photography (a highly scientific number gleaned from my gut).

 

[Louis_Dobson]

The RAW processor as ACR/LR (look at those horrid greens and yellows), and camera one as the E3. Logically, I'd expect camera two to be the D3, but actually I'd expect a much better result from the D3, so maybe not. Mind you, it took me all of ten seconds to stop using LR with the D3, so maybe...

--
http://www.flickr.com/photos/acam

 

[Juergen Hahn]

It was clear from the first set of pictures that the second camera recovered more of the highlights. As far as the second set of pictures are concerned, I am not able to see a clear "winner". Maybe I am looking for the wrong things.

Anyway, thanks for conducting the test.

 

[Big Ga]

It was clear from the first set of pictures that the second camera
recovered more of the highlights. As far as the second set of
pictures are concerned, I am not able to see a clear "winner". Maybe
I am looking for the wrong things.

Oh trust me ... camera two walks it! Its just I pushed both too far.

But the wood is the right colour for a start. The 'purple wood' from camera 1 is simply that the whole of the shadows is comprised of nothing but red and blue channel noise!! I'll try and do a better test tomorrow.

 

[Paul Szczepanek]

I was suspecting it but after seeing the terrible (grid pattern?) noise I'm sure that camera 1 is E3 and camera 2 is D300.

Actualy I don't mind the noise at all, even if it's really terrible but I do mind the terrble lines you get across the frame because they're impossible to tame unlike normal chroma noise.

EDIT:

Also, can you see how the frame of the window leaves a smear on the wood? That's definitely my E3

 

[Big Ga]

EDIT:
Also, can you see how the frame of the window leaves a smear on the
wood? That's definitely my E3

Well I have to say I've never seen the smearing like that before on any oly image (only on cameras with electronic shuttering). It just shows how far I had to push the file.

But I did remember something I have up on pbase - the jar was in the shadows and had 4 stops of exposure compensation applied. You can see how much worse the smaller sensored FZ50 came out.

The E1 isn't shown here, but it was almost identical to the E300 (just a bit less res!).

I'd still like someone to explain to me how come the newer cameras (including the E3) seem to have more noise in the shadows at ISO100 than the E1/E300, yet they are a fair bit cleaner than these cameras once you get up to ISO800 (and I'm not talking fancy in camera smoothing, the E3 files are certainly more usable at 800 than any other fourthirds cam to date, regardless of the converter used) Seems very counter intuitive !

 

[Stainedsilver]

that's a moot point here

Don't you mean "cow's opinion"?

 

[Ehrik]

The extra 'raw headroom' is from colour channels clipping at different
points. How much differently they clip depends on the much the
colour filters mismatch the colour of the light that clips.

Since different cameras often have different relative strengths and
colour response in their Bayer colour filters, different cameras will have
different amounts of headroom in different situations.

One may be better at a cloudy sky clipping whereas another can be better
at a blue sky clipping and a third might be the best one for incandescent
highlight recovery. In theory at least.

Test's like DPR's can give the impression that 'raw highlight headroom' is
a fixed sensor property that doesn't vary with circumstances, when it
is more about how the camera's Bayer filters relate to the studio lighting.
I find it a bit suspect to conclude that one camera is
'better at raw headroom recovery'.

Does this make sense?

Just my two oere
Erik from Sweden
--
CA is a chromatic abbreviation

 

[Jay Turberville]

Since different cameras often have different relative strengths and
colour response in their Bayer colour filters, different cameras will
have
different amounts of headroom in different situations.

The vast majority have a one stop difference in daylight between the red and green pixels. The blue tends to vary a bit. So for the most part, digital cameras all have about one stop of highlight headroom in daylight - assuming you define highlight headroom like you are (and I do).

One may be better at a cloudy sky clipping whereas another can be better
at a blue sky clipping and a third might be the best one for
incandescent
highlight recovery. In theory at least.

Yeah - in theory. But in practice, it is the green to red gap that is usually the largest and I've yet to see it differ by more than about a stop.

Test's like DPR's can give the impression that 'raw highlight
headroom' is
a fixed sensor property that doesn't vary with circumstances, when it
is more about how the camera's Bayer filters relate to the studio
lighting.

Yes. If they aren't using daylight balanced lights, then they could get a different result.

I find it a bit suspect to conclude that one camera is
'better at raw headroom recovery'.

Me too. I strongly suspect that they are all about the same (excepting Fujifilm SR and perhaps Foveon).

Does this make sense?

It does to me. This page has plots of the individual color filter responses for four different digital cameras. They were all shot in daylight. All the values were taken from an undemosaiced (document mode in dcraw) file with no gamma correction. So these graphs show the actual average raw values at one stop intervals. You can see how blue sensitivity varies between cameras, how the one stop difference between red and green is the case with each sensor (Panasonic, Kodak and Sony sensors here) and how the E-330 has two different green sensitivities or perhaps colors.

http://www.jayandwanda.com/photography/DR/four_CFA_responses.html

--
Jay Turberville
http://www.jayandwanda.com

 

[Ehrik]

I remember seeing your document (great test!) last year but it was
good to revisit it in this context. Not dramatic, but clear differences
even though it's only two manufacturers we're looking at.

At that light, eyeballing the curves, it seems the L1 has about 1.15
stops difference, followed by the C7070 at 0.95 stops, E-330 0.8
stops and E-500 at 0.75 stops between the colour channel clipping
first vs. last. So in that test the L1 was "best".

BUT let's say you redid it using the setting sun and blue light was down to
90% and red light up to 110% (normalised to the original intensities and
green at 100%). Then the E-500 might be the "best" one and L1 "worst"!

Which sort of confirms how misleading it can get to try to rate cameras
after how "good they are at raw highlight recovery".

Assuming a largely monochrome scene, a camera that has a highlight
benefit from mismatching colour filters will have a shadow drawback
since the channel that's clipping late in the highlights will be poor at
collecting the weak light at the shadow end. So from a DR perspective
there is a possibility for losing at one end what's gained at the other.

Which I guess can be part of the reason why having some pixels
white (like Kodak is experimenting with) is not a popular idea.

Test's like DPR's can give the impression that 'raw highlight
headroom' is
a fixed sensor property that doesn't vary with circumstances, when it
is more about how the camera's Bayer filters relate to the studio
lighting.

Yes. If they aren't using daylight balanced lights, then they could
get a different result.

And a user using the camera in another situation may not get the
same results. Whatever difference DPR finds between cameras in
this respect will not be generally applicable.

Just my two oere
Erik from Sweden
--
CA is a chromatic abbreviation