May 7th, 2009 by Clint
I don’t have the same pessimism for 3D in the long run as the article that David Newman was responding to with his “No Problem With 3-D”. I have the strong sense that humans are clever creatures, and we will find ways to fix most of the problems that bother us. However, I don’t think we have the technical aspects of 3D mastered. We don’t get nauseous from 3D effects while watching a play on stage, even when Gallagher smashes a watermelon in our face, so I think there is still something afoot.
I can’t yet dismiss the idea that seeking between objects with unmatched convergence and focal properties is a way to cause viewer discomfort, even though David is pretty convincing that we won’t suffer long-term damage. Since the convergence point of the stereo pair should track with the object distance (as David Stripinis has pointed out), doesn’t that limit us to a single plane of interest where the convergence point and focus coincide? That would mean that a close-up large object would have one area that hits the sweet spot, and maybe the further from that area that your eye travels, the more confused you will be.
To revisit a point I touched in “Where Have 3D Stereoscopic Movies Taken You?”, John Batter once discussed how it’s important to carefully pace the amount of depth effect. Poorly paced depth effects lead to increased fatigue, which could also be a nausea contributor.
I’m starting to think that hyperfocal areas don’t work the same with stereo pairs as they do with a single lens (like in David’s optics examples), because the perceived image includes ghosting from opposite eye images. For near objects, I suspect this ghosting dramatically decreases the effective focal depth of a pair of eyes. Maybe this is part of why deep focus establishing 3D shots are my favorite. The amount of blur in areas surrounding objects of interest provide a strong cue as to the expected seek distance to move focus to another object. The camera lenses bake in a single blur amount for the scene, but in real life, the amount of blur changes based on the location of the object of interest with relation to the other objects. I wonder if there is enough of a mismatch between the expected seek distance and actual seek distance to cause a feeling that the world is shifting around unnaturally.
Ordinarily, we shift our heads slightly to “fix” confusing or interfering occlusion and edge problems, but it’s impossible to have a head shift fix anything for projected film. Shifts in object occlusion in real life depend mostly on distance from viewer. If we happen to flinch out of the way of a near-field object flying at us, the projected image won’t shift at all, leading to a definite mismatch between expected and observed motion. This kind of mismatch is what causes seasickness according to my favorite explanation in Alton Brown’s Good Eat’s episode, Rise of the Rhizome (starting at 6:55).
This does hint at another reason for why I love deep-focus establishing shots. Maybe there is only a tiny amount of parallax shift expected for far off objects when I shift my position around, so the lack of shift on the screen isn’t unsettling.
One trick to detect stereoscopic problems would be to track the head movements of a test audience. The more movement for a scene, the more likely it is to be nausea inducing. This might also be related to the classic “audience engagement” metric, where the times when the audience shifts in unison are worthy of extra attention.
I think we should create a 3D film nausea index combining the amount of seeking required between cuts, the magnitude of mismatch between expected and actual seek distance from object to object in a shot, and the ratio of in-focus objects that don’t have matched focal and stereo convergence properties. Once we can quantify the factors that contribute to viewer discomfort, we can automatically warn the editor and maybe someday the director on set, when they’re pushing the limits.
I’m still eagerly awaiting the first 3D gag that breaks the widescreen matte for added effect, even if there is still some discomfort caused by the medium. While I hope 3D can help filmmakers connect with the audience better, all I really want is to be shown another great story.
April 30th, 2009 by Clint
Believe it or not, there is good news to be found on the 2009 swine flu outbreak. This post is a compilation of the news that’s been the most reassuring to me, and contains no bad news or further panicking. Really.
The first and best bit of good news is that many people now know the best ways to keep from catching the flu. Wash your hands often, keep your hands away from your face, and avoid contact with the ill. If we adopt this as a way of life, and not just as a panic response, we’ll see a dramatic fall in cold and flu infections at large.
The next piece of good news is that we haven’t seen a revision of casualty estimates to include recent hospital cases. This is an indication that we’re still in the early stages of the outbreak, which gives us the most leverage for intervention. And so far, as more cases of infection are being discovered, we’re not seeing a corresponding increase in fatalities that matches previous pandemics.
Have I mentioned that I work with some great people? The day before the media picked up on the Flu outbreak, my co-workers had a paper published in BMC Microbiology on work they did last year: “Conserved amino acid markers from past influenza pandemic strains”. Surprisingly relevant, huh?
[Update: An article in Wired, Swine Flu Genes Show Virus May Be Weak, published on May 5, featuring my boss and my co-worker, although the science has been sensationalized a bit.]
[Update: A news broadcast, Computers decipher H1N1 genetic secrets, that aired on June 14 featuring my co-workers. I also make an appearance in the video, sitting at a meeting.]
Which brings us to even more good news: based on the data in Jonathan’s paper, the circulating strain exhibits only a fraction of the markers associated with the worst pandemic strains. While there’s always a causation versus correlation concern, if the markers prove reliable, this may mean the current strain isn’t a worst-case reassortment. As an interesting aside, both the markers and the genome sequencing seem to indicate that the strain is primarily a swine strain with some avian characteristics.
To round things off, you can’t catch the flu by eating pork, and there are still multiple anti-virals that are considered effective at treating the circulating strain.
I hope the good news will keep rolling in, and that we’ll be able to look back at this as a modern model for handling outbreaks.
April 26th, 2009 by Clint
My thoughts are with the victims of the Flu outbreak in the U.S. and Mexico. This is a historic event, and we can expect the news to stay saturated with coverage reminiscent of the SARS outbreak of 2003. This morning in Mexico City, 20 are confirmed dead from the Flu, mass gatherings have been suspended, and schools will remain closed until at least May 6.
You can avoid infection by avoiding close contact with the sick, by cleaning your hands often with soap and water, and by keeping your hands away from your eyes, nose, and mouth. If you are sick, you must stay at home, which you should be in the habit of doing anyway.
There is some good news. The outbreak was discovered much sooner than I expected we could, and that will save lives. To help show how this saves lives, below is a generic epidemic curve.
Disease spreads from person to person up to a peak, and then tapers off as potential hosts are exhausted from the population. Underneath the curve is a second curve, showing deaths caused by infections. The reason I work on disease detection is to try and make these curves as small as possible. The earlier you intervene in an outbreak with treatment and quarantine, the smaller the curves will be.
This Flu strain is H1N1, and in case you’re wondering, there IS significance to the 1’s. This is THE Flu. H1N1 is the type of the 1918 Flu that raged across the world, killed more than twice as many people as World War I in a tiny fraction of the time, and for some reason disappeared from our collective memory. For the “normal” flu that we see every single year, the mortality rate is about 0.1% (1 person dies for every 1,000 infected). A low estimate of the 1918 Flu mortality rate was 2% (200 people died for every 1,000 infected).
Part of the reason so many die from this strain is that one of the most magnificent things in our corner of the universe, the vertebrate immune system, becomes perfectly useless by overreacting so badly that it kills us. Healthy young adults make up the majority of deaths because they have very strong immune systems which can’t stop fighting.
Flu Shy, Don’t Bother Me
There is hope that we can strike down this outbreak like we did SARS. If we succeed, then there’s a chance we’re not too far from finally stopping the “normal” Flu, which every year kills 30,000 people just in the U.S. That’s ten times as many people as perished on September 11, and 75% of the yearly car accident fatalities.
I’ve been telling people I want to make a Flu movie. I imagined it as narrative fiction, and desperately don’t want the opportunity to make it a documentary. We can tell some great stories about love and life centered on the Flu. We don’t need the ridiculous fictitious diseases that haunt theaters. I’m horrified that we accept the Flu as a normal thing, and the coming of it as uncontrollable as Fall or Winter. I hope it goes the fuck away.
February 26th, 2009 by Clint
I can comfortably say that the range of quality of 3D cinema is almost as wide as the range of 2D cinema. Story and celebrity are still king. The introduction of digital 3D cinema seems to be an evolution more like the transition from 2 channel audio to 5 channel audio than the revolutions of sound and color.
The added dimension gives a better sense of scale. It provides another positional queue that can be used to manipulate the viewer’s immersion. Whether two or three dimensions per frame, films are still the filmmaker’s framing of the world through the window of the camera lens.
Technology Can’t Save A Bad Story
I saw a demo of Alioscopy’s auto-stereoscopic (no glasses needed) 3D display, which is a fancy plastic lens on top of an LCD. It actually works, but has a limited number of viewing angles, and requires an array of 8-10 cameras to create the video. I think this necessarily means the effective resolution is one-eighth to one-tenth of the LCD’s native capability.
I also saw a demo of 3D capabilities being built into some new LCD TVs at incredibly low cost. This wonderfully brings 3D into the home, and is claimed to be compatible with Sony’s high-def players. Neither of these techniques are practical for a theater though.
Turtles All The Way Down
The most popular projection technique for 3D is circular polarization, which maintains left/right channel separation extremely well, and is more robust at preserving color levels than linear polarization. The polarization is accomplished with a filter (linear polarizer mated with a magnetic field generator called a retardation plate) in front of the projector. The polarizer, which makes light coherent so the 3D trick works, is the death of brightness, killing over half (~60%) of the projector’s light. The light that gets through is spun clockwise or counter-clockwise depending on the electric field (+ to -, or – to +), which is switching at 144fps, or 6 times per frame. And then he 3D glasses take another 10% of the light. This is all fascinating to be sure, but if you’re switching a single image 6 times per frame to reverse the polarization, you’ve introduced 6 blanking periods.
Blanking periods are fine if they’re part of the media (like in a film projector), but when they’re introduced this way, you get unnatural motion that ghosts and stutters. This is noticeable (and distracting) in any scene with moderate motion – I bet snail documentaries look excellent. One published spec put the filter’s transition time in the blanking period at a very impressive .6ms, which means 3.6ms per frame, or 8.63% blackness as a conservative estimate, but I’d guess it’s closer to 15% in the end. A 2-projector setup isn’t unreasonable to help fix this, and maybe a 6-chip projector will work, but collimating light is a bitch.
A giant LED jumbotron will be even better if we can get consistent color. To make the 3D gag work, each pixel on the screen gets a clockwise or counter-clockwise polarizing filter. If you want to cling to the 4k wagon, then you’ll want an 8k display, but that seems excessive. “Modern” theaters are purpose-built facilities anyway, so why not ditch the silver screen?
“Because I Can” Isn’t Good Enough
3D has hit its stride is with computer-generated films, portraying a synthetic world through synthetic cameras. At The Conversation in Berkeley last year, DreamWorks’s John Batter showed us the Tai Lung escape scene from Kung Fu Panda, specially re-worked for 3D. It was great. John also said they carefully paced the depth effects because it easily becomes tiring, and scene transitions are carefully managed to prevent nausea. One of the more entertaining things I learned from him is that audiences don’t like to get poked, like with the pick-axe-in-the-eye gag that kicks off My Bloody Valentine. With that movie, the break of immersion caused by 3D gags was welcome, and maybe that wasn’t entirely accidental. An oddity worth remarking was the Superman “put on glasses now” scenes that were in 3D – bad idea.
One potential playground for 3D is aspect ratio games. Like the The Dark Knight IMAX (and a handful of other predecessors like More American Graffitti), we’re sure to see 2.35:1 frames being broken for added effect, because the mastering is really 16×9 (caveat projectionist).
The best use of 3D I’ve seen yet was in the awkwardly live BCS championship game. The 3D motion stutter is terrible for watching football, but you can easily grasp the relative sizes of the players in the huddle. One eye of the projection would regularly go haywire, followed by a full audience groan. One shot sang to me every time they used it: a deep-focus establishing shot all the way across the field from a corner of the end zone. The camera played the perfect role, melting away, leaving me wondering why I couldn’t smell the dirt on the field.
I Can Make A Plasticine Pancake
3D isn’t working for me in the claymation realm either. For Nightmare Before Christmas, 3D was added synthetically, and never felt like anything more than 2D cards floating around. With Coraline, the preview was nauseating because of the scene transitions, which luckily wasn’t a problem with the film. Deep-focus shots were again the most powerful, but most of the time the motion stutter was more distracting to me than the 3D was immersive.
Fix It In Post
Seam carving paired with re-projection is a candidate for making live action 3D as workable as computer-generated 3D is. Seam carving is a daunting task in two dimensions, but for many scenes, having stereoscopic pairs will help resolve features. You then rearrange the elements in 3D space, which brings you into the computer-generated arena, where a synthetic world and a synthetic camera are the ultimate in flexibility. The drawbacks to this approach include needing to capture at higher resolutions, and needing lots of light because you want to be stopping down your lens, aiming for the deepest focus possible. The more in focus the entire scene is, the more flexibility you have to adjust the virtual lenses during re-projection.
I Miss Captain EO
So far, story has been secondary to the technology with 3D films. This is almost predictable though, since the technical barriers to 3D are significant. Not only must filmmakers be fluent in the language of film, but they also need to blaze the trail in 3D production, inventing norms as they go. So far, only equipment manufacturers have seriously benefited. If anyone can make use of the immersive potential of 3D, it’s the likes of Steven Spielberg, James Cameron, and George Lucas, with brilliant crews and infinite resources to help tackle their problems. I can’t wait for the great storytellers to tell me another story. If they want to tell it to me in 3D, then so be it, but I don’t think it’s necessary.