15/04 Discomfort

Sweeping statements make me nervous. Daniel Engber's claim in this article for Slate, that "3D" always has and always will hurt your eyes, is a case in point. Mind you, it's always nice to hear some dissenting views about an industry trend with no shortage of enthusiasm associated with it.

I find it interesting that Daniel finds viewing 3D films uncomfortable - an experience that to some extent I share when watching 3D in regular cinemas which use a single projector (rather than one projector each for the left and right-eye views). These projection systems display the two views of each frame sequentially in time, and use a polarising or spectral filter that flips in synchronisation with the projector along with polarising or spectral filter glasses to perform the "separation" that ensures left-eye images are seen only by the left eye (at least in theory) and so on.

Personally, I see a characteristic "shimmer" when viewing such "time-sequential" 3D systems that is most obvious when I move my eyes around the picture. During this kind of "saccadic motion", the human visual system effectively switches off (or otherwise ignores) much of the signal coming from the eyes, in order to prevent you from experiencing a sensation of motion as your eyes move from place to place. If you're watching a time-sequential 3D image, that means each time you move your eyes, they effectively "switch" off and on at times that are completely unsynchronised with the projector, and there may be a perceptible period during which one eye sees no light at all, leading to the perception of flicker in one eye. It's just a personal theory that this is the cause of the "shimmering" I perceive and a contribution to eyestrain, of course, and if any visual psychologists happen to read this, I'd love to chat about it. But the point here is that this is a minor limitation of current projection technology. It's not a fundamental limitation of stereoscopy.

Daniel is wrong to claim that fundamental advances in 3D technology have not been made in the last 50 years though. Film projectors (and cameras) suffer from registration problems: the mechanisms that move the film behind the camera or projector shutter are mechanical, and there is always a certain amount of inaccuracy in the positioning. If the two images in your stereoscopic 3D presentation dance around the screen due to poor registration, that's going to cause major eyestrain. Digital cameras and projectors have no such moving parts, and the single-lens, time-sequential method of projection in particular ensures that the two eye-views are perfectly overlaid without requiring careful calibration. But there are plenty more ways to give people headaches than with poor registration.

For example, as Daniel correctly points out, viewing a stereoscopic image is fundamentally different to viewing a real-world scene, in that the convergence of the eyes and the distance at which they are focussed is decoupled in the former case. This is an extremely well-known issue, and there are a number of ways of dealing with it artistically: keeping things that the audience will want to concentrate on (usually the expensive actors) at a depth close to the screen, for example.

In the long term, I'm not entirely convinced this will still be an issue: anyone who's ever worn glasses has learned to adjust to a different relationship between convergence and the focussing of the eyes - the whole purpose of glasses is to change the way your eyes focus, after all. People with varifocal glasses learn (and usually learn late in life) to cope with lenses in front of their eyes that cover a whole range of focal lengths depending on the part of the lens they look through, which also effectively decouples convergence and focussing.

Stereoscopic film and television is an evolving, immature medium. It's ludicrious to claim that it would be the ideal medium for every genre that we have now (after all, those genres evolved to suit the medium of 2D film and television). But I think it is equally ludicrous to claim that its well-known limitations will lead to inevitable audience discomfort. There are plenty of technological and artistic challenges to be addressed, particularly if broadcasters want to start 3D television services, but I don't think there's a fundamental discomfort issue that precludes the success of the medium.

Oh, and as for the idea that stereoblindness might preclude the success of stereography - did colour blindness preclude the success of colour?

25/09 High Frame-Rate Television

My colleagues and I from BBC Research had a fantastically successful time a couple of weeks ago at IBC, Europe's huge annual convention and conference for the broadcast community. We went there to shake things up a bit, feeling that proposals for improving the quality of television (including High Definition television itself, frankly) had concentrated too much on improving the spatial resolution (the number of pixels) while completely ignoring the temporal resolution (the number of frames per second). Television is, after all, about moving pictures.

By the end of the week, we'd shown our demonstration of the benefits of higher frame rates to hundreds of people, who almost without exception agreed that the improvements were very clear. Our criterion for success was that we'd get higher frame rates on the agenda for consideration in future TV standards, and I think we certainly achieved that. We're looking into what to do next, including finding a way to display images at frame rates higher than 120fps (the upper limit of all the contemporary displays we're aware of* - we've been using projectors designed for alternating-frame stereoscopic 3D so far), investigating how increasing the frame rate of video improves the efficiency of video codecs, experimenting with changing the temporal shape of camera shutters (simulated by down-conversion from high frame rates to conventional ones), and verifying our assertion that shooting at higher frame rates doesn't increase the visible noise in the video signal. Problem is, we really want to do some more work on 3D television, too...

We've published a White Paper on our initial work, available here. It's quite short, contains no maths, and would interest (I hope) technically-minded people from both inside and outside broadcasting. Similarities between the historical section and this post need not be pointed out. :-)

*we have CRT monitors that can go up to 200fps, but not at any kind of sensible resolution.

EDIT: We made it into the "press"! Woo! :-)

13/08 Setting up a (networked) Canon ip4500 on Kubuntu 8.04 (Hardy Heron)

This turned out to be the easiest non-Postscript printer installation I have ever performed on Linux. YMMV, especially if you use a 64-bit distribution.

  1. Download the ip4500 driver software from Canon - I used version 2.80, available here. You'll need both the "cnijfilter-common_2.80-1_i386.deb" and "cnijfilter-ip4500series_2.80-1_i386.deb" files.

  2. Install both these packages, starting with "cnijfilter-common_2.80-1_i386.deb". I got no error messages at this stage.

  3. Restart the CUPS subsystem - from a command prompt enter "sudo /etc/init.d/cupsys restart"

  4. Launch the KDE "System Settings" application and select "Printers".

  5. Select "Add Printer/Class" from the "Add" menu.

  6. Click "Next" and choose an appropriate backend. My printer was plugged into a Buffalo Linkstation on my local network, so I chose "Remote LPD queue" and then entered the hostname and queue name ("lp") appropriately. If you have a firewall installed on your machine, you'll need to enable outgoing lp traffic before doing this.

  7. In the "Printer Model Selection" dialog, select "Other", and then choose "/usr/share/ppd/canonip4500.ppd".

  8. This would be a good point to print a test page, as offered by the printer installation wizard. You may also wish to enable the duplex unit or change other printer settings at this stage, using the "Settings" button.

  9. Click swiftly through the banner, user access and quota dialogs (unless you're special) and then enter a name for your printer - this is what it will be known as on the local machine.

  10. You're done. To set the printer as the default for all applications, right click it in the "Printers" system settigns panel and select the appropriate option.


So far, setting up this printer has been astonishingly painless on both Windows and Linux.