I’ve always found nighttime photography with its deep shadows and bright hightlights challenging. With it is portability and low noise capability, my hope was to achieve some compelling D7000 nighttime photography. The camera is indeed enabling me to do just that, much as the D700 before it, with the premium of its weight and heft, was allowing me. Not only does the D7000 provide terrific shadow recovery, but it does so while retaining vibrant color. Add a proper denoise tool in PP to minimize whatever noise does pop up when we push shadows, and some striking images become your payoff.

Around the Pike Dock and skyline Paker’s and Queen Mary

In a previous article we saw a test case that demonstrated D7000 shadow recovery for exposures at base ISO. From this and samples others around the Internet photo community have posted, we see the D7000′s remarkable ability to pack shadow detail at base ISO (100) with little noise repercusions.

A couple of days ago, while shooting some informal portraits, I ran, entirely by mistake into a situation that allowed me to see how this shadow recovery capability works out in real life. Here are the two original shots, processed through Lightroom 3.4 for sharpening and camera calibration, but without any exposure compensation or fill (LR’s “shadow recovery”).

In this shot, I left ISO at 100 by mistake, and with a shot intended with small aperture and fairly fast shutter speed to reduce the amount of ambient (see bright background), the low ISO resulted in under-exposure.
Realizing my mistake, I hurriedly raised the ISO to 320, but it should have been more like 500-800. The result is better, but still somewhat under-lit.
Here I took the first shot and raised its exposure by 1.2 and fill by 50. The result is what the intended exposure would have been had I thought a little more about what I was doing.
For the ISO 320 photo, I applied 0.3 exposure compensation and raised fill only by 30. Now you can compare this to the previous shot and see which has less noise in the shadows.

Even though the shadow recovery for the first shot yields the least noise, the payoff for the recovery method for the second shot comes in the form of the photo I captured a few seconds later. Still not the best lit shot right out of the camera, but join the D7000′s clean noise performance with the right tools and techniques, and a reward like this might come our way.

There’s been a lot of buzz about the D7000′s outstanding dynamic range at base ISO (100). This buzz includes claims that it bests the full frame D700 at its base ISO (200). To see how these claims would work out in a practical shot, let’s look at the following little vignette, a challenging exposure where we have to decide between preserving highlight detail in the lamp shade against preserving some amount of detail in the deep shadows around it.

On this first set of exposures, we opt to compromise to keep most of the shadow detail, while losing some detail in the highlights.

D700, ISO200, f/8, 1/4 sec

D7000, ISO100, f/8, 1/2 sec

But what if we purposefully select an exposure with two stops two less light? We have now preserved most of the lamp shade detail, but in doing so lose quite a bit of shadow detail, as shown in the first pair of images that follow.

D700, ISO200, f/8, 1/15 sec

D7000, ISO100, f/8, 1/8 sec

However, with shadow protection in ViewNX2 (or an alternative shadow pull technique), we can bring out shadow detail to get the following.

D700, ISO200, f/8, 1/15 sec with shadow protection=65

D7000, ISO100, f/8, 1/8 sec with shadow protection=65

This is not ideal. Neither corrected image is perfect, but it appears the D7000′s output retains the best colors, while the D700 is already falling apart. You may also want to access the full size images to also review the amount of noise the shadow detail pull brings out.

Can we take this one stop further, with more preservation of the highlights at the expense of the shadows? Here’s the uncorrected set of images.

D700, ISO200, f/8, 1/30 sec

D7000, ISO100, f/8, 1/15 sec

And here is the corrected pair of images after applying maximum shadow protection.

D700, ISO200, f/8, 1/30 sec with shadow protection=100

D7000, ISO100, f/8, 1/15 sec with shadow protection=100

I’ll let you decide whether these results are acceptable, but on the face of it, it does appear that the D7000 holds an edge in the shadows at base ISO.

If you have ever dealt with hot sky that led to over-exposed highlights and/or under-exposed shadows and have wondered whether you should run out and buy an Graduated Neutral Density (GND) filter, you might want to see whether an alternative solution is available to you. We have all seen terrific exposures made possible through the judicious and skilled use of GND filters, but GND filters can’t handle every situation or require compromises to be used for some scenes. For instance, if the dark-to-bright line occurs along a jagged mountain top, how would you align the filter’s straight dark-to-bright line? Complicated geometries will require compromise or abandoning the use of the filter altogether. What if we used a post-processing technique that allows us to achieve a Virtual Gradual Neutral Density (VGND) filter?

Let’s look at this next sample image as test case. The first image is the middle exposure for 3-shot HDR bracket, and a compromise between sky and shadow areas has led to some problems. The second image is the HDR composite image: better in the shadows, but still not quite there, and pushing out the shadows (not shown here) any further leads to image degradation. The final image is a straight horizon virtual GND image, where below the buildings we use the bright frame from the 3-shot bracket, and for the sky we use the middle image in the bracket (i.e., the 0EV image), using two layers and a masking technique to blend the two exposures. The dark exposure is used in the background layer and the bright exposure is used for the foreground layer. By selecting and removing the sky from the bright exposure in the foreground, we are left with a composite that includes a brighter foreground (water, sand, etc.), and a well-exposed sky, for a more balanced composite exposure.

Original	HDR	VGND

As you will notice, the VGND version shows more detail in the foreground while preserving good balance in the sky’s exposure. When compared to the HDR version, it also shows less blurring in the surf. And for what it’s worth, this particular image used the equivalent of a 1.3 stop GND. Have you seen any of those on ebay?

Now let’s look at a far more complicated example, a very challenging exposure of a bright early morning scene. Again, the first two images show the original, as-shot middle exposure and the HDR version blending a 3-frame bracket. The third image uses the VGND technique, but with as complicated a mask as you can concoct, isolating the shadow areas in the foreground from the bright areas in the middle and background. Once more, the shadow areas use the +1.3EV frame in the 3-frame bracket, while the midtones and highlights use the 0EV (centered) frame. Combining the two yields nice detail in the shadows in the foreground and calm highlights in the rest of the scene. The only trick here is that your masking technique has to be spot-on, and unfortunately, the author is still fine-tuning his skills in that department.

Original	HDR	VGND

If you go pixel peeping, I’m sure you can find some imperfections in these VGND images, but don’t attribute those to the method or concept behind it. With better masking technique, even better results are possible. Bottom line: when you encounter a difficult exposure, namely one with a lot of dynamic range, take a 3-5 frame bracket, and when you get home, try both the HDR and VGND approach and see which you like best.

Updated Aug 5, 11:20 PST with italic and underlined text.

What is the right exposure? This question comes up often, especially when wondering whether the camera’s meter and automated mode has let us down. As Bryan Peterson writes in Understanding Exposure: “…[exposure] comes up most often as part of a question — a question I’ve heard more often than any other: ‘Hey, Bryan, what should my exposure be?’ … And my answer is always the same: ‘Your exposure should be correct!’”

“Correct,” Peterson goes on to demonstrate, depends entirely on what you are trying to accomplish. In fact, right after this quote he shows a “correct” exposure where the subject, a man, is in shadows but well exposed while the background is severely blown out. Oh, my, what would the Matrix-metering detractors do with such a shot if their Nikon D80 or D90 had exposed likewise?

To illustrate the challenges that both photographer and in-camera meter face, let’s look at an example where, with a much maligned D80, we placed the focus point on the subject, we set the aperture to f/8 and ISO 100, and let Matrix metering give us the shutter speed. We snap the shot, and sure enough, just like everybody says, the background is blown.

Maybe we should spot meter on one of our usual suspects, moving the focus point around until it lands on some of that blue water in the background. We set the shutter speed accordingly and press the shutter, only to find that now the background is well exposed, but the vase is almost a silhouette.

Well, maybe we should spot meter on the vase and see how that turns out. We do that and discover the background is even more over-exposed that in the first photo.

You may start to realize that with this lighting situation and others like it, you, the photographer, will have to make a compromise. How you arrive at this compromise will depend on what is important to you in this image. Do you want a beautifully exposed background? Or is the foreground vase more important? In short, what is this image about?

You could say that it’s about the pretty blues and greens in the background. Alternatively, you could reason that the vase is the subject, and its surroundings, though necessary for the overall look of the image, are secondary. There, you made a choice. Now take it a step further: if the vase is what matters, how do you want it to look? You think about and decide that you like how the back lighting and how it shows off those nice amber hues.

With that choice, it looks like the second shot is the best compromise, but you are not satisfied. The bright background is really too distracting. If the point of this image is to portray the vase, you don’t want anything to take away from it. Do you have any alternatives? You do. One would be to wait until the sun goes down a bit and see if that helps us in the background. As you think about it, though, you realize that the vase will also get darker. That gives you an idea. You have to add light to make this work. If you used the exposure setting for the second sample above, but added some light in the foreground to bring out the vase, how would that work? Here’s one try, with bounced flash to keep the light as natural as possible. You see some improvement, but the nature of the lighting (front-lit vs. back-lit) doesn’t quite give you the look you want.

Another option is to select the first exposure, which thanks to Matrix metering’s inclinations, is over-exposed by +0.7EV in the background, and in ViewNX, pull the highlights down with -0.7EV compensation and push the shadows out with shadow protection. The result is acceptable, and very little noise has crept in the shadow areas.

Still seeking that perfect exposure, you decide to give HDR a try. HDR has gotten a bad name thanks to a slew of cooked up samples, but you are very conservative in how you use it, and achieve the following result with a 3-exposure bracket.

You may have noticed that in this discussion, we started with a question about what constituted the right exposure, which we went on to answer by other means, namely what made this a good photograph. And that’s Bryan Peterson’s point. You can’t know whether you have the “correct” exposure unless you first know whether you have the intended photograph. This is why metering often frustrates us. No algorithm in the world has been programmed with our personal intentions in mind.

As a final observation, perhaps you agree the last two images offer the best balance between lighting and the goal we set out to accomplish in this photo. But it took post-processing to achieve these results, and you may find that distasteful because for some reason you believe every shot should come out of the camera perfect and ready for prime time. I advise you to jettison such a view-point. When the lighting is tough, you will have to use all the tools in your kit, and you better have post-processing as one of them. In fact, with experience you may find that as you approach shots like these you will start making them with post-processing in mind.

By now you probably have tinkered with your DSLRs histogram to validate the correctness of your exposures (if not, see the references at the end of this blog entry), and you understand what pushes a histogram to the left or right. It’s the exposure, right? If you have set your camera shutter speed, aperture and ISO correctly, or applied the right metering and exposure compensation, you will capture the right amount of light, and your histogram will be a beautifully centered, non-clipping mound. Right?

As it turns out, though photography is indeed about capturing the light, it turns out other related factors drive your histogram. Even if you shoot RAW, the histogram your camera displays belongs to the thumbnail JPG it produces — not the actual RAW data. You may see clipping, but is the RAW data really clipped? To answer this, we need to understand that the R,G,B histogram is driven by color settings on your camera — how it interprets and converts the RAW data into a final JPG file, or the JPG thumbnail included in the RAW file (NEF for Nikon DSLRs).

This brings us to 4 in-camera settings that affect color rendition, and hence, histogram distribution: White Balance (WB), Sharpness(!), Saturation, and the effective tone curve. In a Nikon D90 and other cameras with equivalent features, the last two are encapsulated in a Picture Control. In a D80, equivalent settings are achieved through the Picture Mode settings. Let’s take each of these in turn, using ViewNX for ease of demonstration, to show how they push and pull your histogram.

First, let’s take a look at the following sample shot, with the in-camera settings intact. As you can see, its histogram may not be the typical “perfect, centered hump” I referenced above, but we see no clipping of highlights, so we’re very happy.

Now, let’s suppose you had applied some really extreme sharpening in-camera. We’ll emulate it here in ViewNX, but something very similar would happen in the JPG’s histogram you would see in the camera’s LCD. As you can see, you would see some clipping in the reds and greens — not because suddenly the scene got brighter, or because Matrix metering blew the exposure, or because you erroneously set the shutter speed one click too slow. All that changed here was sharpening.

Next let’s suppose that instead, you shot the photo with the Vivid Picture control. Again, we see clipping, but this time in the blues. Would you blame Matrix metering, try to apply negative compensation, or beat yourself up for having the blinkies? But again, nothing changed in either the exposure or the scene. Only what the camera is doing to the image has changed.

Now let’s play with White Balance. Taking the Vivid shot, let’s warm up the White Balance to 5800K from the default 5200K for Direct sunlight. As we would expect, the blues recede from the highlights, the reds advance toward the highlights, and we get a better behaved histogram that though still clipping in the highlights (now for the reds), does so minimally and for most practical purpuses, acceptably.

Finally, let’s show what could happen if we take a photo with the wrong White Balance. Going back to the original as-shot photo, let’s suppose we mistakenly used a left-over temperature setting for a previous photo we took earlier in the day (I know this happens to no one but me, but play along for a second). That temperature setting, around 4100 K is too cool for this particular scene, so, sure enough, we get accentuated blues that clip in the highlights.

That’s all very interesting, you may say, but what conclusions can you draw for practical application? Here’s a few suggested take-aways:

1) When judging exposure in-camera, keep in mind that the camera’s color and sharpening settings are influencing the histogram you are reviewing.

2) Setting the right White Balance, or getting as close to “real” as possible in the field is essential if you want to ensure proper exposure. Even if you subscribe to the “I will fix it in PP” strategy for WB, keep in mind that if you get the wrong exposure because the WB you had in the field led you astray, you may get good White Balance in post-processing, but you may not achieve a well exposed image.

3) Even when shooting RAW, if you intend to apply a given Picture Control in post-processing, it is very useful to have that same Picture Control in the camera at the time you shoot and review the histogram so that you can accurately evaluate the exposure.

4) The effects of in-camera sharpening are not as pronounced, but do play a role, and in conjunction with the right/wrong Picture Control or WB setting may also lead you astray. Even if you sharpen in Post-processing with Unsharp Mask, keep in mind that it will shift the histogram. Depending on the situation, you may need to give yourself some headroom to avoid clipping.

5) Finally, if you play with colors in post-processing, such as hue or saturation changes, you may also need to give yourself some headroom to avoid clipping in the final post-processed image.

References for further reading:
Understanding Digital Camera Histograms: Tones and Contrast
Understanding Digital Camera Histograms: Luminance and Color