For those who shoot video on a regular basis, high contrast lighting situations generally provide a challenge, as you have to make a trade-off between correctly exposing for the shadows or the highlights in the frame. Thanks to HDR video techniques you can have the best of both.
In an earlier post here on ReelSEO.com back in September, fellow-author Christophor Rick talked about HDR for video. This article builds up on that, albeit from a videographer's perspective.
HDR Video - or High Dynamic Range - already is quite common ground in stills photography. Thanks to the bracketing-function on most any DSLR (read my DSLR how-to guide here), you can instruct your camera to take sequential exposures at several stops below and above what is generally regarded as correct exposure. After that, you can combine these images inside Photoshop to a HDR new image that is capable of showing both detail in the shadows and the highlights of the image and have them correctly exposed.
On the left you can see that the image was underexposed. The image in the middle does the opposite - blown out skies, but better exposure for the water. By combining these two images inside Photoshop (using a process of tonal mapping), you can create an HDR image (on the right) which exposes correctly for both of the high contrast situations.
Unlike our eyes, cameras have a very limited contrast-ratio, typically anywhere between 30:1 or 50:1. Contrast that with the human eye which is capable of handling ratios that vary between 400:1 to 10,000:1 and you'll quickly understand why cameras cannot record what we see in real-life. With HDR and 32-bit float technology, we are capable of upping the game.
From Stills to Motion: HDR Timelapsing
One can easily imagine the above procedure working for several of these HDR photos in a sequence, more commonly known as a timelapse. By instructing your DSLR to take a bracketed exposed picture at an interval basis (say every 10 seconds) you can create something like the video below (please select 720p playback):
Although the effect is beautiful, timelapses are generally used for special effect purposes in a typical video production. For example, take those fast flowing skies over the New York skyline in most any episode of The Apprentice, inside the video above or these amazing TimeScapes from Tom Lowe. As timelapses span longer periods of time, if you include people, they move at high speeds. This can be a desired effect. I frequently use it inside my own productions as build-up or transitioning shots. However, timelapses (even HDR ones) cannot be used as a technique with live talent in realtime. The contrasty lighting situations, however, do occur from time to time, so all we need is to find a solution to this problem.
Caveat: True versus faux HDR
Before we move on to the possible solutions, one important note must be made. Most video cameras are not capable of recording RAW footage (yet). Currently, only a few cameras support RAW recording, including RED One/Epic and ARRI Alexa. Both of these companies are said to hold a race for getting a patent on an in-camera solution of recording well beyond 12 stops of latitude in a single take. RED states that using the new HDRx technology, the latitude can go up 18 stops, while ARRI mentions a mere 13 stops of latitude. Look at this clip released by RED - although it appears dull at first, that very flatness contains tons of data that can be manipulated in post production. Data that would normally be lost due to either too much contrast/sharpness in the picture and/or high compression.
It is widely expected, however, that Canon will at some point in the future unleash a clean HDMI output for their current or at least their next breed of DSLRs (hopefully native RAW recording) — this is, after all, the number one request from users. When connected to an external recorder such as the Ki Pro, nanoFlash or Ninja, even the much anticipated Panasonic AF100/101/102 (depending on where you are on this planet) can record at much higher datarates and with less compression, thus preserving more of the fine detail.
Therefore, until we have RAW recording more widely available, the solutions below may be described as faux HDR. When compared to the world of stills photography, faux HDR is the same as creating a HDR image from two JPEGs, instead of RAW files (true HDR). RAW files, of course, record so much more data of the light captured by the sensor than what remains after it has been translated into a compressed JPEG (8-bit) file format, as Andrew Kramer shows in this tutorial.
Regardless of whether it is true HDR or faux HDR we are talking about here, the general idea is to significantly increase the latitude of the images recorded by correctly exposing for the highlights and the shadows. Instead of realising that in one take (as with the RED or the ARRI), we have to resort to increasing the overall latitude by mixing multiple recorded takes: essentially, by making a combination of under and overexposed areas. Much like we do with creating HDR stills.
There are two methods to this end: either sequentially or simultaneously.
Sequential Recording: Motion Camera Control
Basically, this technique involves the shooting of a backplate and live action take sequentially.
For example, an actor which is positioned directly opposite the sun or against a window. Common camera sense would advice against such high contrast situations (e.g. never shoot against an opposite light source — a lesson I learned before changing my career towards camerawork), but sometimes there is no other choice, or it is a deliberate creative choice. Under normal circumstances the above mentioned trade-off comes into play. However, with motion camera control you can sequentially correctly expose for both the lighter and the darker areas of the frame. Motion control can be achieved by either locking off the camera or repeating the same move.
Typically, you would lock off the camera on a tripod: make sure you lock off the camera completely, which probably means using a remote start-stop device to control the record button. Then, first expose for the background (the window) and record for the same period of time as the live action takes place. Next, you redo the scene, but now with the actor in place and you expose correctly for his face. As a result the window will most likely blow out (a white blob). In post production, you can then combine this 'backplate'-take and the sequential 'action'-take into a single HDR video. As far as I know, there is no software or plugin available to do this easily, but through a combination of rotoscoping and/or tonal mapping, I would imagine this to work (read on until you see the video at the bottom of this article).
However, locking off the camera typically results in quite static shots. In order to make your shots more dynamic you could use a pan/tilt/dolly move. But when doing things sequentially, requires you to have identical, repetitive camera motion. Although there are very professional solutions available, nowadays there are also more affordable ones. An example of this are the motors and controllers made by Kessler Cranes. Although these were originally designed for their PocketDolly and CineSlider products, a more recent addition - the ShuttlePod - allows for travel over longer distances.
The beauty is that Kessler Cranes also supplies an Oracle controller that allows for repetitive motion, meaning that you can repeat the same camera move time and again. In the second part of the video, you can see how Tom Guilmette superimposed himself at multiple takes of the same camera move. It is not hard to understand how this technique can be used for sequential HDR video recording as well.
The upside of this solution is that you can do this with a single camera unit. The downside of the sequential approach, of course, is that lighting situations change, especially when shooting outdoors. Think of the sun moving behind a cloud, or the speed at which the sun sets or rises. That's when you need to look out for the other solution. Here's another video from Timelapses.es, who uses a similar motorised camera control system:
Simultaneous Recording: 3D Beam-Splitting Technology
Currently, there are no cameras available yet that have built-in beam-splitting technology, although I have been told that ARRI holds a patent on some technology to that extend. Basically, recording HDR simultaneously means using two identical cameras/lenses.
Although I am not a believer in the recent surge of 3DTV technology (I consider it more push than pull, all supply and no demand), I can totally imagine 3D technology taking off for HDR video purposes. The acquisition methods of beam-splitting cannot only be used for creating images suited for 3D, but also help solve a real-world problem in 2D — being that of high-contrast lighting situations.
In 3D there are two methods of recording - either by placing the cameras side-by-side, or by using a mirrored solution of two cameras being placed in a 90 degree angle. Both videos below from 3D Film Factory explain the differences of these two techniques:
Placing the cameras side-by-side provides a desired effect for 3D stereoscopic movie recording. Much akin to how are eyes work, the side-by-side placement creates the illusion of depth. However, for HDR video purposes, this set-up tends to create more problems than solve them, as you get a parallax effect which cannot be easily solved in post production. Parallax is an effect that you might want to strive for, as this camera set-up of 52 DSLRs so nicely shows, but when it comes to HDR prevention typically always is better than the cure. So the beam-splitting solution is the way forward for true HDR video recording, as evidenced by the videos below, from E3D Creative and Soviet Montage:
The upside to beam-splitting is that you record an identical image at the same time, meaning that there will be no differences in lighting situations or time. In order to create an HDR video, all that needs to be done is expose the cameras differently. In our example, one for the background, the other for the actor. Subsequently, you combine them in post production (as explained before).
The downside clearly is that you need both a beam-splitting rig and two sets of identical cameras/lenses in order for it to work. This instantly triples the gear needed for a production, but thanks to the DSLR revolution, gear is becoming more and more affordable. Nonetheless, for now this approach easily involves $10k worth of equipment.
Conclusion: 3D's real future is HDR Video!
HDR video definitely has a future, as it solves a real-world everyday problem for videographers and cinematographers alike.
Unlike 3D, I like to add. In September I visited the International Broadcast Convention (IBC) in Amsterdam (more coverage on that later) and everyone and their neighbor had solutions for a market that is nacent at best. Don't get me wrong, I believe 3D is fantastic for feature length movies inside theaters — we have all seen Avatar. Epic. However, I just don't see me and my family huddling around the TV set each wearing stereoscopic glasses. Not now, nor at any point in the future. I just don't see it happening. Now I know that some manufacturers are developing solutions without requiring you to wear 3D glasses, but these solutions are far from perfect. Above and beyond all this, I just bought a new FullHD TV set which I expect to last for at least 5-8 years. And I think I am not alone. Look up the statistics and you can easily predict that 3DTV has a long and winding road ahead. All supply, virtually no demand.
Nonetheless, as explained in this article, the technology behind 3D, however, can be put to excellent use for creating moving images with a HDR look an feel to it. Agree or disagree? Please feel free to comment below.
Want to see more HDR videos? Feel free to watch/join/add to this dedicated HDR Video group on Vimeo.