Report - Researching 3D

Almost 50 members of UK Section and students met on 17th January at Ravensbourne College to hear about Research in 3D.

Chris Johns, Chairman of the UK section, in calling the meeting to order pointed out that we were meeting just 50 year after the death of Sir Isaac Shoenberg – who led the dream team of EMI engineers, including Alan Blumlein, who in 1931-35 had developed Electronic Television leading to the Launch of the BBC’s High Definition (405 line) TV service in 1936 – 2D of course, and we were going to hear about and see how 3D worked – the adding of another dimension to our viewing.  

Dr Lizzie Jackson, Academic Development Manager for Ravensbourne’s Broadcast Production and Engineering courses, welcomed SMPTE to Ravensbourne – and outlined Ravensbourne’s role in providing the industry with Broadcast engineers and other media professionals, and being a centre of research in media and its use of technology. In particular Ravensbourne was launching a new BA in connected Television – probably the first in the world.  They were seeking funding under the European Union Seventh Framework Programme to extend the works started by Martin Uren and his son James on Stereoscopic 3D (S3D) – and are looking to SMPTE members to help them.

Martin Uren, Subject Leader Broadcast Technology at Ravensbourne, and a well-known industry training consultant, started his presentation by reflecting on how when he was younger he was drawn to 3D in the cinema which was providing something emotional that you could not get from 2D representation. This is something that was very complex as it was not just the physics/optics but was deeply linked to how the human vision system perceives the world.  This as at the heart of his practice-driven research methodology – exploring what actually happens and adapting the aims of the research depending on what is discovered.           

He introduced the basics of human vision - stereoscopic vision comes when we fuse images from our two eyes to create a feeling of depth – but given the human visual acuity and interocular distance these two images are essentially identical for anything over about 50 metres distant.
In stereoscopic vision we actually rely of a large numbers of non-stereo cues – Motion Parallax, Occlusion, relative size being the most obvious – but linear perspective, texture gradient, contrast elevation and aerial perceptive are all used.  We also have feedback from the muscles that rotate the eyeballs to converge on the selected object – and to focus the eyes’ lenses.  It is these which are in conflict when you focus on the screen and converge on objects in front of the screen – i.e. with negative parallax

In the real world we are immersed in our environment – so, as S3D is a more immersive viewing experience, if any one of the depth cues is broken by the S3D effects – the viewer is thrown out of the immersive environment – and thus the “reality” that is being created.

Finally there is stereopsis where the two images presented to the brain are slightly different – and being able to see round an object – and shadow stereopsis (here light and darkness can make an object appear concave or convex “is what you see a crater or a mound??”)

Martin demonstrated many of these effects using a 3D camera pointed at the audience – and particularly what happens with an object close to the camera – and how this could look very strange/ different when it popped out of the screen. He pointed out that as we get older our accommodation decreases – but Stereo 3D relies on us not being too confused by our eyes focusing on the plane of the screen while they converge on the object. This can be a strain if has negative parallax - it is a lot close to you than the screen.  This vergence/ accommodation conflict is what makes S3D unnatural for all viewing - S3D can only give an illusion of depth – unlike Holographic methods which are true 3D.

From the fundamentals of how we see – to the Immersive experience – Martin pointed out the differences between the environment of the Theatre – where you view a large screen some distance off which you and others in the theatre see with almost the same viewpoint with the far peripheral vision being all black, and the home TV viewing where you are close to the screen to get the same field of view, or further away so that the rest of the family can view with you and the room is unlikely to be black. 

Martin, in conversation with the audience who had many questions and points from their experience, then outlined some possible guidelines for making good S3D - 

The “don’ts” – steer clear of zooms, fast cuts, unsteady cameras, dissolves – avoid edge violations, Depth jumps, soft focus foreground objects – and never pull focus.

And the “Do’s” – have plenty depth of field, use slow camera moves, use diagonal  framing to enhance depth cues (this may mean shooting from a lower position) – use lighting and action to take the viewers’ attention to the action, enhance all monoscopic depth cues – allow time for the viewer to explore the scene….. And smoke or slow motion is very effective.

All of these were only guidelines – and they could be broken for effect – but the viewer needs some recovery time with good but less extreme S3D.  But the illusion of S3D should not be broken – by making the stereo cue not consistent with an all the other monoscopic cues.

These guidelines showed that the Grammar of S3D production was likely to be different from that of 2D. He observed that much of this has not been learnt yet as there has been so little “Real not CGI” S3D production across many genres. His research was designed to increase the knowledge of S3D in various genres by actually making short items which highlighted the issue and likely best practices and grammar.

He also pointed out that it was impossible for one person to do an impartial score of a comparison between a S3D sequence and the equivalent in 2D, and raised the important points that those working in the production may be more tolerant of S3D artefacts – for instance excessive negative parallax, than the first time viewer. Martin likened this to having a new pair of spectacles (eyeglasses) – as you leave the optician the world looks very strange and different - but after a day or two you do not notice any strangeness as you put on or take them off.

All of this showed the flexibility of the human vision system and that there is a lot more work to be done on how we see and how the S3D cinematographer can exploit this to make even more immersive experiences for the audience.

Martin Uren ended with an account of the experimental productions he has made in collaboration with other members of Ravensbourne’s 3D research community -

Starting with a S3D pilot for a shopping channel – which is a close space for the “performance” - but involves close ups on the items being sold. He pointed out that to get the depth of field required to do a close up on Jewellery required the cameras being stopped down, thus requiring a lot of light, and the flares off each facet were different in each eye. This gave a very unnatural 3D effect – whether the effect was due to camera position or the polarising effect of semi silvered mirror.

After light – A short film which used chromakey to add a 3D image within a 3D image. We viewed this and marvelled at the way in which the keyed images had depth and matched the S3D depth of the foreground image – see a 2D representation  and “the making of

Love You To Pieces – where Martin was one of the actors – so he observed the Crew at work from the other side of the lens – a Red based mirror rig.

Pave The Way – a documentary about a Dorset artist and his relationship with the landscape – shot on a side by side rig for the landscape shots and a unibody camera for the close ups.

All of this was part of the practice driven research promoted at Ravensbourne – and he announced the research team’s next project – capturing the Northern Lights aurora in north Norway in S3D. The team have estimated that the camera interaxial distance would need to be at least 40 km! (25 miles).

As well as the storytelling and S3D image issues Martin mentioned some of the practical problems – for instance Pro-sumer cameras turning themselves off to save battery – even when they are in record and mains powered – and that the workflow is very important – he was often having to decompress the camera files because they could not be put into a professional system format and work flow. That required a lot of storage, took time and lost quality.

Martin well exceeded the normal duration for a lecture – but had kept the audience captivated with his exposition and the ways in which he answered questions all the way through – drawing out the issues of S3D Production and Storytelling and offering practical guidelines.
These notes cannot cover all the things that were discussed by a very active and knowledgeable audience.

Despite the lateness – members and students talked for an hour afterwards – carrying on discussions started before and during the lecture –while consuming Pizza and sandwiches.

SMPTE UK Section thanks Ravensbourne (www.rave.ac.uk) for the venue & opportunity to meet students and Weitzel.tv for the generous quantity of refreshments.