By David Wood, EBU
The route to ‘First Generation’ 3D television (S3D) requires journeys down many roads. These include the development of broadcasting formats for different circumstances where different degrees of existing and new equipment are needed by the viewer. These are being considered in the DVB Project as ‘Phase 1 3DTV’ and ‘Phase 2 3DTV’.
As well as this, there is much work to be done on production technology for 3DTV, including developing rules for production grammar that will provide programmes that are most comfortable to watch. We also need agreement on file formats and other aspects for production. The EBU 3DTV Group, led by Andy Quested, BBC, is studying these issues.
But at the same time as these tools for production and broadcasting are developed, everyone working in 3DTV needs a general understanding of the factors that affect the quality of perception of 3DTV. This article is a short introduction to them.
The science of the perception of images is termed ‘psycho-physics’, and our question is thus, for the psycho-physics of 3DTV, what is important – what matters?
In a 2D television environment, we have only to concern ourselves with picture quality evaluation itself – sharpness, resolution, colour fidelity, degree of artefacts, etc. The ITU has well defined methodologies for evaluating the ‘subjective’ quality of TV images.
In a 3D television environment, we have a lot more to concern ourselves with. New ways of evaluating 3DTV systems are needed, which may include, but also go beyond the classical subjective evaluation methodologies used for 2DTV. These have not yet been formalised, but ideas of how to do them are emerging. For example, the author proposed that quality evaluation of 3DTV should be made based on subjective evaluations using 50” screens at five times picture height. There are reasons why this is a good combination, because it removes some possible elements of bias in the testing. There may be other (and better) ways of breaking the psycho-physical issues down to manageable portions, but one way is to consider the problem in three chunks, but which are all inter-related to an extent:
• The realism (‘quality’) of the image
• The eye discomfort associated with using the system, and
• The degree of eye fatigue that the system causes.
If we look further at these three individually, we find a number of subelements which influence them.
For our first area, we can find five factors which influence the quality of realism of the image (assuming the programme maker’s intention is to achieve maximum realism). These are as follows:
• The basic quality of the Left and Right eye images themselves. All of the elements that affect 2D images come into play. It matters whether the image (using the ITU acronyms) is ‘LDTV, SDTV, EDTV, HDTV, or UHDTV’. The 3D quality will get better as we step up the ‘2D quality’. It is because we have HDTV today that 3DTV becomes more attractive than it was in the SDTV and EDTV days.
• The potential depth resolution of the 3D images. S3D systems fold depth information into the Left and Right eye pictures by virtue of the disparity between them, and thus the depth resolution is determined by the horizontal resolution of the Left and Right images. It matters whether there are 720, 960, 1280, 1440, or 1920 samples per line; and, all other things being equal, the potential depth resolution increases with each step.
• The geometrical congruency of the objects in the scene. This is the relation of the width/height/depth of objects. Quality is affected by the relationship of width to height to depth of objects in the 3D images compared to those in real life. Do they look the same as they do in natural vision? Do they look ‘right’?. Or are they like cardboard cut-outs, or people with ‘octopus arms’. How ‘right’ they look is determined by factors such as camera lens angles.
• The apparent size of objects in the scene. This is the extent to which their size matches normally perceived sizes. Do they appear as ‘very small men and women’ – the ‘puppet theatre effect’, or the opposite – the ‘baby’s eye view of the world’. This is mostly related to the spacing used between camera lenses, and how far it is from the normal human eye spacing, which is about 6.5 cm.
• The construction of the scene. This has an influence on the ‘impact’ of the scene, and thus on the perceived quality. An ‘ideal’ 3DTV picture (at least some argue) may have some central object which has elements that project just forward of the screen plane, but which do not ‘violate’ (cut) the TV screen window.
For our second area, we find three subissues that define the level of eye discomfort. These are as follows. As explained earlier, complete separation of the issues is not possible and there is mutual influence of factors.
• Ease of fusing the Left and Right eye images in the mind – ease of accomplishing the process technically known as ‘stereopsis’. Our brain takes in the two images, and lays them on top of each other in the mind. It creates a single image with depth or volume – or at least it will do so provided the two images are adequately related to each other. This is affected by the amount of disparity or parallax between the Left and Right eye pictures. It is also affected by where the objects in the scene lie with respect to the ‘screen plane’ and the depth range. More generally, this fusing ability is affected by the goodness of the matching of the Left and Right eye pictures. One of the most critical issues can be vertical misregistration of the L and R images. Possibly the precision of time coincidence of the Left and Right images may also be a factor.
• Comfort level associated with glasses. The weight, elegance, level of light attenuation of the glasses affects the 3D experience.
• The production grammar used in making the programme can affect eye discomfort. One need is for the depth plane of the principle object in the scene not to change dramatically on shot/scene cuts, and the extent to which the principle objects are placed in the 3D ‘favourite’ location or ‘sweet spot’, which is at, just forward, or just behind the screen plane.
For our third area, we come to that for which there may be least knowledge available today. This is the sensation beyond eye discomfort, which might be called eye fatigue. Eye discomfort can turn to eye fatigue, which is something experienced after viewing for a time. There can be mild nausea, or a kind of seasickness, or irritated eyes. They can be due to many of the factors which are described earlier in the article. They may lead to conflicting balance cues. The old enemy of accommodation convergence conflict, where we try to focus on the ‘apparent’ position of the object, can be a contributory factor. One of the important questions that we need to answer is whether there may be residual effects of watching 3DTV.
So why do these all occur? At heart it is because S3D in NOT natural vision – it is a limited ‘subset’ of it. The S3D camera (like all cameras) does not record the phase of the light, just its amplitude. The ‘phase’ of the light wave we see in 3DTV is the phase of the TV screen, not the original objects. No focus depth cues are available, and only one pair of views are available, wherever your head is. We can never ‘lie down on the settee’ and watch 3DTV, even wearing the glasses. It only works when the viewer is sitting or standing vertical.
There is much research to be done on the psycho-physics of 3DTV, and until we know more it may be sensible for broadcasters to urge viewers to take off the glasses if they feel the slightest eye discomfort, and possibly not to watch for long periods without having a pause without glasses. 3DTV can be a very exciting experience, but let’s take care.
David Wood is the Deputy Director, Technology and Development at the European Broadcasting Union.
Copyright Tech-i 2011. Published with kind permission of EBU Technical.