27.12.2014 Views

DEliverable 2.3 - the School of Engineering and Design - Brunel ...

DEliverable 2.3 - the School of Engineering and Design - Brunel ...

DEliverable 2.3 - the School of Engineering and Design - Brunel ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

ICT Project 3D VIVANT– Deliverable <strong>2.3</strong><br />

Contract no.:<br />

248420<br />

User Acceptance Validation Plan<br />

especially for live events, because, as is <strong>the</strong> case for stereoscopic 3D, here <strong>the</strong> cameras must also be<br />

exactly aligned <strong>and</strong> synchronised.<br />

Even though auto-stereoscopic displays do not <strong>of</strong>fer a large parallax area, <strong>the</strong> viewer appears to be<br />

able to look around objects. However, <strong>the</strong> number <strong>of</strong> viewing positions <strong>and</strong> <strong>the</strong> resulting stereo pairs<br />

is limited. In this case, <strong>the</strong> shear distortion effect is reduced but still detected in each stereo pair.<br />

Never<strong>the</strong>less, if <strong>the</strong> viewer moves around in front <strong>of</strong> <strong>the</strong> display, jumps in <strong>the</strong> picture can be detected<br />

when <strong>the</strong> viewing angle changes. This phenomenon is called flipping. A fur<strong>the</strong>r disadvantage <strong>of</strong> <strong>the</strong><br />

auto-stereoscopic process is <strong>the</strong> missing vertical parallax. As <strong>the</strong> views can only be separated<br />

horizontally, no vertical parallaxes can be displayed. The result is that a viewer can appear to look left<br />

<strong>and</strong> right around an object but not over or under it.<br />

More recently, a combination <strong>of</strong> conventional 2D video capture with depth map generation has been<br />

used for <strong>the</strong> capture <strong>and</strong> processing <strong>of</strong> multiview auto-stereoscopic 3D content. However, <strong>the</strong> display<br />

<strong>of</strong> multiview auto-stereoscopic 3D content relies upon <strong>the</strong> brain to fuse <strong>the</strong> two disparate images to<br />

create <strong>the</strong> 3D sensation. A particularly contentious aspect for entertainment applications is <strong>the</strong> human<br />

factors issue. For example, in stereoscopy, <strong>the</strong> viewer needs to focus at <strong>the</strong> screen plane while<br />

simultaneously converging <strong>the</strong> eyes to different locations in space producing unnatural viewing<br />

(Yamazaki et al., Lambooij et al. 1989). This can cause eye-strain <strong>and</strong> headaches in some people.<br />

Consequently, content producers limit <strong>the</strong> depth <strong>of</strong> scene to be viewed to minimise this problem. With<br />

recent advances in digital technology, some human factors which result in eye fatigue, such as limits<br />

in head movement in <strong>the</strong> case <strong>of</strong> circulate/linear polarized glasses systems, etc., have been eliminated.<br />

However, some intrinsic eye fatigue factor, like a mismatch in convergence <strong>and</strong> focus, will always<br />

exist in stereoscopic 3D technology (Onural et al., Benton, Honda 2006). Fur<strong>the</strong>rmore, due to <strong>the</strong> lack<br />

<strong>of</strong> perspective continuity in 2D view systems, objects in <strong>the</strong> scene <strong>of</strong>ten lack solidity (cardboarding)<br />

<strong>and</strong> give rise to an ‘unreal’ experience.<br />

For <strong>the</strong> 3D video quality assessment <strong>of</strong> stereoscopic videos, most researchers employ subjective<br />

testing (De Silva et al., Hewage et al., Leon et al.2010) focusing mainly on depth perceived by <strong>the</strong><br />

users on autostereoscopic displays <strong>and</strong> <strong>the</strong> sensitivity <strong>of</strong> <strong>the</strong> observers to <strong>the</strong> changes in depth in a 3D<br />

video scene (De Silva 2010). Most <strong>of</strong> <strong>the</strong> 3D video user perception studies relate to <strong>the</strong> design <strong>and</strong><br />

evaluation <strong>of</strong> 3D stereoscopic <strong>and</strong> multiview video systems based on different coding parameters. In<br />

<strong>the</strong> majority <strong>of</strong> <strong>the</strong>se studies, subjective testing using a qualitative methodology with no more than 15<br />

users has been employed (Kalva et al. 2006; Saygili et al. 2009; Knorr et al. 2008; Olsson <strong>and</strong><br />

Sjostrom 2010). In some <strong>of</strong> <strong>the</strong>se research studies, different types <strong>of</strong> video systems were compared by<br />

<strong>the</strong> users in terms <strong>of</strong> various system parameters <strong>and</strong> user perception <strong>of</strong> stereoscopic versus multiview<br />

video (Knorr et al.2008; Olsson <strong>and</strong> Sjostrom 2010). Most <strong>of</strong> <strong>the</strong> results suggest that, for stereoscopic<br />

<strong>and</strong> multiview video, <strong>the</strong> bit rate <strong>and</strong> <strong>the</strong> content <strong>of</strong> <strong>the</strong> original 3D image form <strong>the</strong> factors that most<br />

significantly affect <strong>the</strong> perceived 3D image quality (Kalva et al. 2006; Knorr et al. 2008; Olsson <strong>and</strong><br />

Sjostrom 2010; Reis et al. 2007). In terms <strong>of</strong> multiview 3D video, it is also noted that users prefer less<br />

apparent depth <strong>and</strong> motion parallax when being exposed to compressed 3D images on an autostereoscopic<br />

multiview display (Olsson <strong>and</strong> Sjostrom 2010). Fur<strong>the</strong>rmore, it was found that motion<br />

<strong>and</strong> complexity <strong>of</strong> <strong>the</strong> depth image have a strong influence on <strong>the</strong> acceptable depth quality in 3D<br />

videos (Leon et al. 2008).<br />

2.<strong>2.3</strong> Data Collection Methods<br />

The following data collection methodologies will be employed to ga<strong>the</strong>r information on <strong>the</strong> users’<br />

perceived quality <strong>of</strong> <strong>the</strong> project’s video content:<br />

• Questionnaires – Comprising <strong>of</strong> both closed <strong>and</strong> open ended questions.<br />

• Interviews – Semi-structured (prepared <strong>and</strong> spontaneous questions) interviews comprising <strong>of</strong> a<br />

selected user sample <strong>and</strong> one/two interviewees.<br />

01.09.11 11

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!