Fullscreen HTML5Med Res (??MB)HTML4

Breakthrough Aussie innovation . . . MAKING 3D MOVIES While there has been a lot of publicity around the latest introduction of 3D movies and the accompanying 3D screens and glasses, 3D movie production is very expensive. In fact, it comes as a surprise to most people to find that the blockbuster “Alice in Wonderland” was actually shot in conventional 2D and labouriously converted to 3D later! Now there is a new Australia camera rig, the SpeedWedge, which promises to streamline the whole process. By BARRIE SMITH siliconchip.com.au June 2010  21 T he blockbuster “Avatar” set the new standard for 3D movies and film-goers have been very enthusiastic. Part of its success is due to the very good 3D camera work but the bulk of the 3D cameras has been a major disadvantage. The set-ups demand that the paired cameras either employ prisms or partially reflecting mirrors to permit a controllable separation of the two lenses, to capture the left and right image pairs. The average human eye separation is around 65mm, so image capture is best served by the camera lens’ interocular distance (IOD) set at about 70mm for most subject material. For close-ups, a smaller IOD is preferred. In “Dial M for Murder”, often recognised as one of the best of the 1950s’ 3D movies, Director Alfred Hitchock was forced to use a large and inflexible camera rig. In one key dramatic scene he used a scaled up telephone to provide an extreme close-up. The reason: it was physically impossible to rig the lenses to give a closer IOD. A recent Australian innovation, the SpeedWedge, could make things much easier. It was developed by physicist and stereographer Leonard Coster. The rig consists of a housing that holds a pair of gen-locked Silicon Imaging SI-3D digital cameras. One camera is placed on top, its lens pointing downwards and aimed onto a partially-silvered mirror with 50% reflectance. This camera captures the left eye image. Beneath it is another, matching camera installed horizontally within the rig, its lens pointing ahead and looking through the same partiallysilvered mirror. This camera captures the right eye image. The complete rig is mounted onto a television camera tracking pedestal. Fig.1 shows the general concept. The partially silvered mirror is the key, with each camera receiving half the light from the scene. By having the cameras mounted at rightangles to each other, their effective lens separation can be varied from zero to as wide as is desired, without any mechanical interference between them. In practice, if the scene involves action in the foreground, the IOD is set to a small value. Conversely, if the scene or subject is more distant, the IOD is set to a large value. While any video cameras could have been used in the SpeedWedge, the Silicon Imag- Realising that the 3D camera setup would not capture macro shots, Director Alfred Hitchcock organised a scaled-up phone for a key scene in the 1950s movie “Dial M for Murder”. Bwana Devil is a 1952 drama based on the true story of the Tsavo maneaters. It started the 3-D boom in the US filmmaking industry from 1952 to 1954. Fig.1: the over/under Speedwedge arrangement to hold the two cameras: the upper camera captures its view via the 50% reflectance mirror, while the lower, horizontal camera is aimed through the mirror’s 50% reflective surface. 22  Silicon Chip ing cameras were chosen because they use the SiliconDVR recording software which solves a major post production problem. SiliconDVR records the two camera data streams in one go so, in terms of the capture workflow, a major task is handled elegantly. As Coster says: “If you can’t synchronise your cameras and record the two data streams eas- Fig.2: the IOD is the distance between the axes of the two lenses and the convergence distance is the distance from the camera to the object they are both pointing at. Some people also refer to this as the convergence angle, which is the angle between these two axes. siliconchip.com.au ily on set, you’re in a lot of trouble!” The SI camera heads have a significant advantage with their small size. This leads to a complete rig that can be picked up by one person. If you wanted to strap two big film cameras into the housing, you could do it but the weight and final size would be impractical for hand-held operation. In practice, the Speedwedge rig allows the IOD (inter-ocular distance) to be varied from zero (for macro shots) to 70mm, covering mid-range and telephoto shots. In practice, the IOD needs to be set differently for each and every scene and the actual setting depends on how strong the director wants the 3D effect to be. This leads to a further consideration. If a wide angle or telephoto lens is to be used in 3D shooting, what it does to the apparent depth in the scene has to be taken into account .Telephoto lenses tend to fight the 3D effect because, even in 2D photography, a telephoto lens gives a fore-shortening or flattening effect. Hence, the perception of depth is quite poor with a telephoto shot. Leonard Coster says it is much the same in 3D shooting: “We can try and push a little bit of apparent depth back into it by increasing the interocular separation. However, you have to be very careful that you don’t produce too large an offset in those images’ background and foreground divergences on screen — otherwise you make the vision too hard for your viewers.” He stresses “we’re not producing a perfect reproduction of the real world because we may not be using the same size sensor and a standard focal length lens or ‘normal’ IOD, but I want to give the audience a comfortable stereo Leonard Coster the and beam-splitting camera rig. On-set checking of the stereo effect can be made with a display set up as an anaglyph (red/cyan) picture or with a crosspolarised monitor, viewed through appropriate specs. siliconchip.com.au June 2010  23 The Speedwedge rig used on another recent production, also photographed by DOP Tom Gleeson and directed by Tahnee McGuire. image that is immersive and visceral without causing eye strain.” Data handling As already noted, the SI-2K cameras’ data streams are recorded to two hard drives. Coster adds that if you’re using other broadcast cameras, the data may go to flash memory cards or a hard drive; if you’re using 35mm film cameras it goes onto two film rolls. On-set monitoring can be accomplished by using a video display, with the pair of left/right images shown on screen as an anaglyph (red/cyan) image. It is viewed through the familiar red/cyan specs, just as you would a 3D movie. Alternatively, you can use small cross-polarised monitors which take two colour signals and give you, with polarising glasses, full colour stereoscopic viewing. For post production, you still edit, Coster explains, just as you normally would, with two streams of vision for every scene. You can merge these two separate data streams later, then have a file which represents a single series of frames, in which there are a myriad of post production paths. It’s even possible to create a stereo DCP (Digital Cinema Package) file that allows you to deliver a hard drive to any DCP-compliant cinema in the world. The cinema operator can load it into the server and play back the 3D vision through the house projectors. In the post-production process, overall colour corrections and convergence can be adjusted. The latter process involves off-setting the two images right or left relative to each other. What this effectively does is rack the entire set back and forth, determining what the audience will see at the screen plane. The shoot Colour grading and adjustment of convergence can be made post shooting, thanks to Silicon Imaging’s software. 24  Silicon Chip Producer-Director Bernie Zelvis was asked by SMPTE Sydney in 2009 siliconchip.com.au Converting ‘Flatties’ to ‘Deepies’ It may come as a surprise to some to find Tim Burton’s spectacular 3D movie “Alice in Wonderland” was not originally shot in 3D but photographed in 2D. The same applies to “Clash of the Titans”. Hollywood producers are now looking through their back catalogs to find suitable titles that can be converted from 2D to 3D, to cash in on the current fervour for 3D titles. The last two Harry Potter films are likely candidates as are classics, such as early Star Wars, Titanic and other major titles. The result is not always a perfect transformation: many viewers who saw Clash were, to say the least, unimpressed, with one blogger claiming the film was “flawed in so many ways, not least because of its underwhelming visual appeal, its lack of ‘3Dness’ but also because the story is just as flat as the visuals.” The “Clash” conversion is reputed to have taken 10 weeks to perform at a cost of around $US4.5 million. to produce a 3D short for its Dimensionale 3D film festival. To do it, he called on Leonard Coster to supply his new 3D rig to be used by Director of Photography (DOP) Tom Gleeson. The result was Highly Strung, with a running time of two minutes. Tom Gleeson admits that, like many, he had a dim view of 3D “based on cheezy movies and red/blue paper glasses.” Then, for the first time, he viewed an HD film on a 3D cross-polarised monitor: “I was at first astounded and then converted. HD and 3D are a potent mix.” In his view a large part of a DOP’s job is to create a sense of depth in 2D images using lighting, lenses and composition. When confronted with images that actually have depth there needs to be a rethink! He says that, after a lifetime watching, analysing and creating 2D, it can be confronting when the paradigm shifts. Once immersed in 3D shooting you have powerful new tools like IOD and convergence that control this new depth. Gleeson recalled that the footage looked “amazing, with depth that felt like you could walk into it.” On one occasion he used smoke to enhance the lighting in the shots and to help create a sense of volume. He feels 3D can immerse a viewer within a picture and story like no other format can. siliconchip.com.au The US company involved was Prime Focus who developed the software and employed an Indian facility to do the actual leg work. How is it done? A process called rotoscoping is at the heart of it: using part manual and part ComputerGenerated Imagery (CGI) processes, an operator hand traces the main elements in each scene, so separating them and allowing each object to be tracked and “converted” to produce the second eye’s view. This of course can be extremely complex, depending on the scene. The first step is to separate the shot into somewhere between two and eight layers of depth. One example may be an image of a person standing in front of a building, with a blue sky and clouds behind. The operator can separate this shot into three layers: the person, the building and the sky with clouds. Contour lines are then drawn around ob- jects in each layer and a topographic layout created with depth lines to indicate the position of each object in the stereo window. Naturally, the objects might well be moving in the succession of frames: computer software can track this movement and create ‘in-between’ frames, so avoiding the laborious effort of tracing each frame. The software also assesses and inserts detail that may be behind each moving object. At this point you have a collection of objects that may look like cardboard cutouts situated at different depth planes. To ‘round them out’, texture maps are taken from each object and overlaid on the shapes. This will give each character facial depth, costume detail, etc. Coster remarks: “It’s a lot of work. Ultimately, it can produce very good results but it is not as good as shooting in 3D. However, you do end up with a movie asset of far more value.” In Hollywood, the dollar speaks loudly. Leonard Coster has created an iPod app which helps handle the stereo configurations when a cameraman is on a 3D shoot. To convey and enhance the depth in each shot he kept the camera movements fluid, with long tracking shots. The cameraman must also give consideration to editing: fast cutting 3D shots can be challenging for viewers, so longer and wider tracking shots are often more suitable. Cutting points also need to be thought out, as parameters such as convergence and IOD should not jar. 3D Drama Although the SMPTE short was Bernie Zelvis’ first foray into 3D stereoscope production, he currently finds himself writing a TV drama series specifically for 3D. From the exercise, he discovered the 3D viewing experience can be easily ‘broken’ by trying to squeeze too much space range onto the picture and ends up hurting the eyes. He concludes there are “traps for young players”. In his opinion, a stereographer like Leonard Coster is necessary to keep you from trying things that just won’t work, as well as to supervise the post process. “The biggest plus with the system we were using was that completed shots could be projected onto a screen, only minutes after the shoot: This impressed all who saw it. One TV executive said this wasn’t even on their radar... it is now!” SC June 2010  25