Pentax K20D Report—Part III: Color, Tone: New Features

By Carl Weese

Last time we looked at some test results working with RAW files. Of course not everyone likes to work with RAW. The K10D offered two flavors of JPEG: Natural, and Bright. While not quite Baskin Robbins yet, the K20D has expanded the line of JPEG flavor offerings considerably. These are accessed from the Function button (not the general menu) and include: Bright, Natural, Portrait, Landscape, Vibrant, and Monochrome. Mono lets you select modes that mimic filter effects: None, Green, Yellow, Orange, Red, Magenta, Blue, Cyan, Infrared Color. You can punch in warmer or cooler toning of the monochrome image and customize contrast and sharpening.

In each of the color modes you can alter the starting point and punch in your own changes to Saturation, Hue, Contrast, and Sharpness. The various modes are in themselves simply pre-set suites of these settings. Natural looks good to me, especially if the Contrast setting is backed off one or two clicks. In fact that looks very close to the ACR 4.4.1 interpretation of the RAW file on my defaults (which use slightly lower black point and contrast settings than Adobe’s defaults). If I need to shoot RAW+JPEG in order to have immediate access to non-RAW files, that’s the setting I will use. All of the others are variations on the theme of “how I don’t want my pictures to look.” Hyped-up candy-apple color. To someone else of course these may be nice bright colors that look just right.

Going back to that white-clipping test on the RAW files, the Natural JPEG mode will mimic the results of that test, but without the benefit of ACR’s Recovery function. That is, there should be full detail through 2.3 stops above the middle, but then clipping will set in and be quite severe at +2.7. So these JPEGs will have less retention of highlight, less reach into the high values than RAW files even on default. The Pentax Photo Lab software will observe these settings in presenting you a default view of a RAW file, but of course since it is RAW you can make any changes you want; the camera settings serve only as a starting point. The other JPEG flavors will reduce the tonal range even more—this is the tradeoff for the questionable benefit of snappier contrast and higher saturation. Happily, none of these settings have any permanent effect on the RAW files, so if you want to play with these controls I recommend setting the camera to RAW+. I admit I’m curious to see what happens with the monochrome settings and will report when I’ve had a chance to play.

Here’s a really novel feature that I’m only going to describe briefly because it would take a lot of testing and experience to get a handle on it. The K20D offers “expanded dynamic range.” What’s that, you say? Let’s look in the manual: “By using the Expand Dynamic Range function, you can expand the light level expressed by the CMOS sensor pixels, making it more difficult for bright areas to occur in the image.”

Right….

To turn it on, you go to the ISO settings through the function button, then hit the function button again to toggle from normal to “D-Range 200%.” When you do, ISO settings below 200 become unavailable. JPEGs shot with d-range on show a big increase in ability to hold apparent highlight information. I say apparent because, while there’s an enormous increase in how far up you can expose without clipping (farther than a RAW file on defaults) there’s something funny about the color information. Like using a massive amount of Recovery in ACR, this appears to be extrapolated rather than direct data. Another serendipitous accident: when I first ran a highlight test with d-range on, I neglected to reset a manual WB setting from a previous test. This resulted in a slightly yellow color in the test run. The normal JPEG files showed this warm bias from mid value up to +2.3, above which they clipped. The d-range JPEGs held off clipping all the way to +3.7, but the color had lost its yellow cast, gone nearly neutral, by +2.

Pictures of actual subjects consistently show an effect on JPEG files from the d-range feature, but the effect varies greatly depending on the subject and brightness range of the scene. In high contrast situations like backlight or harsh cross-lighting, the feature can give a dramatic improvement. On more normal subjects it can sometimes give a strange HDR-like appearance. Mid-tones are generally raised, but shadows are almost completely unaffected. In many cases the blending effect reduces color saturation. If you want to play a real game of concentration, you could try testing those jazzier JPEG flavors like Bright and Vibrant combined with d-range.

Here's the clutter next to my imaging computer, harshly lit by a desk lamp. The histogram from a JPEG shot on Natural  shows clipping of the highlights (this low rez view hides how high the right side of the histogram goes up the wall).

The 72 ppi sRGB screen grab of a shot at the same settings but with d-range turned on probably won't show much improvement on your monitor, but look at that histogram. There is a major improvement in the highlights and upper middle tones, though no real change in the shadows.

Here’s what I think is going on. They call it 200%, and you lose the first stop of normal sensor ISO range. I think the camera is doing a form of HDR on the fly, combining the basic data of a shot at ISO 200 (for example) with the highlight detail that would have been captured at ISO 100. The effect is tricky. It’s quite difficult to get exactly the same look by adjusting a RAW file in ACR, although it’s easy to get a similar, just-as-good-or-better result. For anyone who has to shoot JPEG files it would be a good idea to get a handle on this feature because the improvements in some situations are very real. But I don’t think it should be used as a “set it and forget” default option. One last point on d-range: unlike all the other picture flavors offered by the K20D, this one affects RAW files. Not anywhere near as much as JPEGs, but RAW files shot with the feature turned on hold a little more highlight detail than with it off. You can avoid clipping at +3.3 with less, almost no, Recovery. The effect is even slightly different for PEF and native DNG captures. So it looks as though both the sensor and the JPEG engine are participating in this game.

Getting smarter
Now on to some reactions to actual picture taking with the K20D. The first thing I noticed examining folders of files in Bridge was that the camera’s automatic white balance has gotten a whole lot smarter. Accurate AWB is vital for shooting JPEGs since substantial color correction of an 8-bit file is bound to cause quality losses. With RAW a bad AWB just forces you to fix it, there’s no loss of quality (an exception is if you were setting your exposure using a tri-color histogram and the WB was so far off that, once corrected, you get clipping in a channel that was unclipped at the wrong WB). But it’s nice to open a folder and see a slew of consistent captures without spending the first part of a work session changing the temp/tint settings of every other exposure. While on this, a nice touch is that the camera lets you fine-tune the AWB. The default setting was too green and too warm for my taste, but through the function button it was easy to open a color box and move the hot spot left and down to adjust the color to my preference. ACR tracks this change and shows the adjusted color as long as WB is left at the default “as shot.”

Spinos Pawn Shop, Waterville, Connecticut

(Auto white balance, auto (program) exposure on multi-segment metering, DNG on the defaults in ACR run through my usual web-posting batch action.)

The next thing I noticed was that the multi-segment pattern metering also seems to have had its IQ raised. Exposure compensation is needed less often, and when needed, you don’t have to use as much. I’m actually having to re-train myself to dial in less compensation with the new camera. Of course the fact that there is a bit more overall dynamic range than offered by the K10D may be helping the meter by giving it a broader target to hit.

Next installment will be after I’ve had a chance to do a lot more actual picture taking.

_______________________

Previously:
Pentax K20D Report Part I: Focus
Pentax K20D Report Part II: The CMOS Sensor

_______________________

Carl

Carl's website 
Working Pictures
WPII

Comments

[Matthew Miller]

In the K10D, the Bright and Natural tone curves are much more than just preset values for Saturation, Contrast, and Sharpness. They change the gamma curve for the resulting image, making quite a difference. There's no "Hue" option in the K10D, but it doesn't sound like that replaces the curve exactly either, so I expect that this is actually the case in the K20D too.

[Ctein]

Dear Carl,

Again, what is the total exposure range for the camera?

Without knowing the shadow-side range as well, I can't determine the import of the highlight side range, because I don't know where the meter has set the exposure point in the total exposure scale.

pax / Ctein

[mcananeya]

Re: Expanded dynamic range

Carl,

This is probably just the same thing that other manufacturers are doing with their cameras. Here is a description from DPReview's review of the Fuji S100fs:

"It would appear that - put simply - the S100FS is under-exposing the shot by one or two stops (to preserve the highlights) and applying a custom tone curve pull the other brightness levels back to their correct values.

Another way of looking at this is to think of the highlights as being shot at ISO 100 and the shadows at ISO 200 or ISO 400 since the camera is effectively applying a non-linear gain to the raw output (which inevitably looks very dark if opened in a non-Fuji development application). In effect this isn't that different to the options offered by Nikon (D-Lighting), Canon (HTP) or Sony (DRO)...

You don't get something for nothing, of course. Shooting at 400% [200% for the Pentax] limits you to at least ISO 400 (hence giving an exposure two stops shorter than the camera's base ISO setting), and this does mean the shadows get noisier, though to be fair the results aren't bad at all. If you can live with the additional noise, then it's a useful feature."

See: http://www.dpreview.com/reviews/fujifilms100fs/page16.asp

Best regards,
Adam

[Hugh Crawford]

Does the “D-Range 200%.” only work for JPEG or does it work in raw as well?

I keep hoping that a camera company will come out with a way to change the linearity of the sensor gain, which would work for RAW as opposed to highlight and shadow masking in JPEG processing which would be nice for people who shoot JPEG , but not much use for me.

BTW, anybody know why tweaking the sensor gain curve at the time of exposure isn't a feature? It really doesn't seem like that difficult a problem.

[Carl Weese]

Matthew, changing the "contrast" setting does in fact change the shape of the gamma curve. Looking at the histograms, I can see it is not analogous to the PS Contrast dialog but more like the contrast slider in ACR, which steepens or flattens the slope and changes the shape of the default curve in a complex manner. Combine this with a shift of Hue and another in saturation (plus a difference in sharpening which will affect the perceptual contrast) and those are all the controls needed to make big changes in the look of the JPEG.

Ctein, as noted in a reply last time, in RAW on default there is still recognizable tone/detail (textured vs smooth target) at -5.7 stops (it reads level 4) though I doubt this tone could be separated from black by any printer. The jpg files seem to track the RAW quite well on the low end. That number is at ISO 100. I'd expect noise to show up with d-range turned on since the ISO setting gets bumped up. To say how far into the shadows it can go needs a judgment call on what level is too black to print, combined with a judgment call on what amount of noise is acceptable, and another on how well, say, detail exposed -2 can be raised to middle. Since I don't have any intention of shooting JPEGs this isn't going to get on my todo list. But the d-range feature definitely does not make the shadow retention any worse, so at modest ISO, there should be at least 5 stops below the meter's middle that at least show clear separation from black and can image a textured surface. On screen.

Adam, I've done some pretty severe rescue work with everything ACR has to offer on files that had to be underexposed to avoid clipping the highlights. There's a look to files that have been manipulated this way. I've also experimented with Adobe's implementation of HDR in Photoshop CS2 and 3 and there's a distinctive look to that (which is similar to HDR examples I've seen from others using different software). The Pentax JPEG files created with d-range look to me much more similar to HDR than to curve-hammering. I don't know what's going on under the hood, just what I see in the files.

Hugh, as noted, there is a slight improvement in highlight clipping point, but it is interesting more because there is *any* impact at all on the RAW file than because the amount of impact is meaningful.

[Chris Combs]

To be pedantic: Those histogram screenshots cannot be directly compared. One has the /!\ alert icon, which means that it's "fake," computed from incomplete data. Clicking the icon will recreate the histo from the full image data.

[Charlie]

What in Green Acres are you all talking about?

[Carl Weese]

Chris, not pedantic: technically accurate.

Setting up the damn screen shots from the two files so they'd be viewable in a blog post was not easy. They went through numerous iterations, and at some point one lost the step of clicking to make the hist kosher. For the record, the change in this case is infinitesimal and couldn't be seen in the reduced view anyway. But thanks for pointing out the tech error.

[Matthew Miller]

Carl — I don't mean that those settings don't affect curves, but that I don't think it's possible at all to replicate one tone curve mode simply by adjusting the sliders in another one. In other words, that setting does more than just set default levels for the other options. (Much more.)

[Michael Arcilla]

As a matter of fact, there *is* a slight difference between the two photos, Carl. It might just be my monitor, but your mentioning the highlights in the histogram drew my eye to any possible blown highlights. And there it is, the highlight at the top of the stack of packing tape rolls. The effect is slight but you can clearly see where the detail is saved (the green copy printed on the core of the roll.)

[Carl Weese]

Matthew, I checked again, and you are right that the various modes are doing more than you can get at through the customize settings. Based on the color diagram helpfully provided in the LCD dialog, what they are doing has more to do with saturation and hue (varied saturations and hues for different colors, like playing with ACR's HSL panel) than with the tone curve. These turn out to be "color palette" adjustments as well as gamma curve adjustments.

Now, if only they would offer a "Restrained" mode, we'd be getting somewhere.

[Colin Work]

The expanded range sounds very much like the highlight protect feature in my Canon 1DIII - and I guess I've come to pretty much the same conclusion. It works in so far as it prevents blown highlights (I figure I can get 2 - 2.5 stops additional overhead), but the price is a degree of unpredictability in the final image quality.

I've played with it a lot, and have decided that it's not acceptable as a default setting, but can be very useful in some conditions - fast action in changing light for example, when there just isn't time to dial in the correct exposure compensation.

It's a sort of security blanket - better to do without it, but when things get tricky, and you only have one chance to get the shot, it may just save the day.

Some people won't like it, some would prefer it wasn't included - but no one says you have to use it, and leaving it disabled has no impact on 'normal' picture taking

Cheers

Colin

[Ctein]

Dear Carl,

Thanks for patiently repeating yourself; I missed that information in the comments in the previous column.

So if I understand correctly, there's about three stops range above the exposure point and six stops below. By my standards, I would call that overexposing (as argued in a previous column of mine).

The question that would leave in my mind would be whether the camera is doing anything that I couldn't do in my RAW conversion. I don't know.

I suppose a way to find out would be to make two photographs of a step tablet, one exposed with the d-200 mode and the other "under"-exposed by a stop in normal mode. Pull them both into Photoshop and create a curves adjustment for the latter that matched the tones of the former. Then do some pixel peeping and see if there was any difference in image quality in any of the steps.

If there wasn't, then all the camera is doing is saving me the trouble of creating a custom profile and applying it en masse. Nice but not earthshaking. If there was, that means the camera is doing something cleverer than I can do, and that's a valuable thing.

I believe the Fuji S100FS has a similar feature; when I get my review unit I'll try to run this experiment and see what happens.

Understand that I'm not pooh-poohing this feature; anything that improves average picture quality is a good thing. Just trying to figure out if it's merely nice or really important.

~ pax \ Ctein
[ please excuse any word salad. MacSpeech in training! ]
======================================
-- Ctein's online Gallery http://ctein.com
-- Digital restorations http://photo-repair.com
======================================

[Carl Weese]

Ctein,

Thing is, the camera's middle lands right around the middle of the 255 levels. (Not talking Pentax here, any good digital camera.) But it retains about 5-6 stops of subject luminance in the 0-127.5 levels, and barely 3 stops from there up to level 255. So what's middle? We could dial in -2 compensation to place the range of the scene within the outer bounds of the sensor's reach, but then all the things we want to have middle tone will be underexposed. Grass, blue sky, all will be much too dark and will have to be raised--a lot--in value. Bad enough to take something from Zone III to Zone V in a RAW file, pretty awful with an 8-bit JPEG.

For a long time we got along with the fact that b&w negative films only held full detail for two stops below the meter reading (less if you used the manufacturer's ISO rating) but with carefully tested development easily held detail four or five stops over. Shooting large format film developed in pyro and intended for platinum printing, I can expect to hold full detail at -2, some detail but definite tone separating from black at -3, but on the other end of the scale expect easy retention of bright values at +5 or +6 and above. It's not symmetrical. Color negative film also shows asymmetrical retention of highlight values.

As I've said before, digital capture just turns this asymmetry upside down. I'm much more comfortable with the way negatives do it. That just may be from 40+ years experience with negs and just a few years with digital capture, but actually I think lopsided-to-the-highlights is more like the way we actually perceive things.

[Murph]

I am a Nikon shooter, but I am just tickled pink that Pentax has come out with such a winner of a camera. Cameras like the K20D make things better for all of us, by raising the bar, and forcing everyone else to stay honest. Bravo Pentax, and keep it up!