This is the first installment of what may turn out to be an irregularly presented, informal series of posts dealing with high dynamic range (HDR) imaging. I don’t really have a firm plan for the series, so it will probably present itself in fits and starts. I have no intention of creating a comprehensive treatise on the subject but I have a series of thoughts that are generally related to HDR that I’d like to express.
I’m starting with the working assumption that everyone reading this has a decent grasp of a working HDR definition; (very) briefly, we’re talking about the assembly of multiple exposures of the same subject via some sort of “tonemapping” software, with the rough goal of obtaining a better, richer tonal frame than can be obtained from a single image that isn’t subjected to a tonemapping process. Obviously, there’s subjectivity to the evaluation; what constitutes “better,” for instance, is a matter of opinion.
Having dispensed with the preliminaries, it’s time to jump into today’s installment–in-field technique, i.e. how to best go about the process of securing the frames that will make up an HDR set, followed by a recent set of experiences in my seemingly endless quest to establish an effective seamless in-field workflow.
I’ve been fiddling around with HDR, in one form or another, for nearly 10 years now and during that period of time the best (there’s that word again) set of options for establishing a baseline of exposures to assemble the HDR set has been something of a moving target. The goal, however, has remained unchanged. When obtaining the images for the set of exposures the minimal considerations involve accomplishing the following:
1) acquiring a set of exposures to completely cover the entire tonal range in a given scene–to ensure that detail has been retained everywhere in the image, from the brightest highlights to the darkest shadows;
2) accomplishing step #1 with as little camera movement as possible (to minimize any issues aligning the multiple exposures in the tonemapping process);
3) doing the above as quickly as possible.
Why is it important to proceed as quickly as possible? Unless you’re working with an entirely static scene (i.e. none of the elements are subject to movement and the light is controlled/unchanging), the more quickly you complete the process the less likely you are to have any problems with alignment or (forgive me for getting technical) “funky lighting.” :)
Some technique specifics that have always been optimal:
1) Use a tripod. The rationale here is obvious–since aligning multiple frames is an inherent part of the HDR process, keeping the camera static is important, if not imperative. Besides, for the kind of subject matter that generally lends itself to HDR, using a tripod is always a good idea.
2) Use a remote release and autobracketing. This addresses point #2 in the the previous list. Using a remote–or the self-timer means that you don’t have to manually press the camera’s shutter release. (Similarly, using an exposure delay means that even if you do press the shutter release on the camera, any movement so introduced will be dampened by the time the shutter itself opens.) Using autobracketing means that you don’t have to touch the camera to adjust the shutter speed. Autobracketing also speeds up the process of producing the set of exposures.
For some inexplicable reason, autobracketing with Nikon camera bodies has always been limited to increments no larger than 1 EV. Most of the time I use a bracketing sequence with the D800E, given the excellent dynamic range of its sensor at base ISO, of four stops. (Sometimes I’ll fire off exposures covering six or even eight stops, but it’s usually five (and essentially never any less than that.) Because of the 1 EV increment limit, that means five exposures (at meter, -2 EV, -1 EV, +1 EV, +2 EV). There are times when being able to simply take three shots (at meter, -2 EV, +2 EV) would be preferable. I actually prefer the processed look of the one-stop increment sequences (I’ve done a lot of comparisons), but firing off three shots takes less time than five and there are occasions when a three-shot sequence could be shot without elements moving whereas a five-shot sequence cannot. It would certainly be nice, in any event, to have the option; it would take no more than a firmware tweak to enable this capability and I have absolutely no idea why Nikon has never seen fit to do so.
For me, the biggest problem with regard to completing a satisfactory exposure set has been implementing mirror lock up. In case you’re unfamiliar with the way an SLR camera operates, the optical viewfinder allows the photographer to see “what the lens sees” via the use of a mirror. When the shutter release is tripped, the mirror is flipped up, the shutter opens for the appointed amount of time, then closes, and the mirror flips back down. When the mirror is raised, the entire camera vibrates to a certain extent (exactly how much depends upon the camera and lens in use). Typically, the impact of this vibration on very short exposures and very long exposures can’t be detected, but in a mid-range of exposures (often in the rough range of about 1/30 second to 1 second or thereabouts) mirror-induced shake can be detected. A careful examination of the resulting image will elicit decreased sharpness due to mirror-induced vibration.
The way around this problem is to lock up the mirror; all Nikon camera bodies have an easily implemented mirror lock up feature. In this instance, using a remote release, one click of the release and the mirror is locked up; wait a couple of seconds for the vibration to be dampened and then click the release again, which will open the shutter. This is fine for single shots but for a set of exposures the problem is obvious–initiating this process for, say, five exposures is going to take a lot of time–probably something on the order of 20 seconds, depending on the specific shutter speeds chosen as part of the sequence: click the release to raise the mirror, wait two seconds for the vibrations to disperse, click again, exposure is taken, click the shutter to raise the mirror, wait two seconds, etc . This is a rather evident violation of the “quickly as possible” rule above. Unless there’s literally no movement in any of the elements of the scene, this won’t work.
When I first came upon Live View–when I bought the D700 back in late 2008–I thought that I might finally have the mirror lockup problem licked. With Live View, the mirror is, by definition locked up. I figured that, using a combination of a continuous shooting setting, autobracketing, a remote release and Live View, that I could initiate the sequence and match all of the three criteria in the first list above. Given the use of a remote release, continuous shooting and autobracketing there would be no need to touch the camera and the sequence of exposures would be taken as quickly as possible. With Live View, there would be no issue with mirror slap either.
Or would there?
With the D700, yes there would. For some reason, which I simply do not and never will understand, the D700 Live View feature worked like this. The mirror was raised, and the image would appear on the camera’s rear LCD. The optical viewfinder was blacked out (because the mirror was raised). The shutter release was pressed…and the mirror was lowered, then raised again, then the shutter opened, then closed. Yes, you read that correctly…the already raised mirror (remember, for Live View to function at all, the mirror must be raised), was lowered, then re-raised. There is no mechanical reason why this should be the case; goodness knows there’s no logical reason why this should be the case either. I have never heard a good (or bad, for that matter) explanation as to why the D700’s Live View worked this way, but it did…and it meant that my would-be solution to the mirror lockup problem for bracketed exposures was no solution at all.
When the D800 series of cameras was released, I looked into the details and determined that this Live View eccentricity, present in the D700, had been corrected. With the
D800/E, the mirror operated in Live View in the manner you’d expect: the mirror was raised, and the image would appear on the camera’s rear LCD. The shutter release was pressed…the mirror remained up, the shutter opened and closed. With this method in place, I figured, the problem was solved. In Live View, using continuous shooting, autobracketing and a shutter release, everything would operate quickly and smoothly. I was experimenting with it just the other day, under controlled circumstances. I preset everything the way I wanted, pressed the shutter release on the remote in locked setting, the five-exposure sequence fired off quickly and…I waited…and waited…and waited. I didn’t regain control of the camera for 31 seconds. It took that long, apparently, in Live View mode to write all of the images to the SD card I was using. When I tried the sequence without using Live View, there was no blackout period at all. Shooting single exposures in Live View there was no blackout. Evidently the camera’s buffer is non-functional in Live View mode. It was a combination of Live View and the five exposures, I reasoned, that caused the 31-second blackout. The images were produced exactly according to expectations, but a 31-second blackout afterward? Every time? That was no good.
The D800 series of cameras takes two types of flash memory–Secure Digital (SD) and Compact Flash (CF). My experiment above was using an admittedly sluggish 32 GB Sandisk Ultra card (rated at 30 MB/s). When I tried running the exact same Live View autobracketing experiment when writing to a rather old 16 GB Sandisk Ultra CF card (also rated at 30 MB/s) the end of sequence blackout period was 7-8 seconds. That’s quite a difference–and something I could live with. I presently have a 32 GB Sandisk Extreme Pro SD card (rated at up to 95 MB/s) on order; when it arrives, I’ll put that card through the test and see if this allegedly much faster SD card will give me performance with multiple Live View exposures similar to (or, even better than) that experienced with the old CF card. But however it plays out, it appears that I’ve finally settled upon a technique that, with the proper media available, will produce HDR exposure sets in something at least approaching a usable method.