University of Surrey

Test tubes in the lab Research in the ATI Dance Research

Non-reference depth map quality evaluation in immersive video applications.

Haddad, Nasser (2016) Non-reference depth map quality evaluation in immersive video applications. Doctoral thesis, University of Surrey.

Nasser_Haddad_PhD_Thesis_Final.pdf - Version of Record
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (5MB) | Preview
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (165kB) | Preview


Three Dimensional Television (3DTV) and Free Viewpoint Television (FTV) are emerging multimedia applications, which promise to offer a new Quality of Experience (QoE) dimension. In order for such applications to offer an immersive experience for users a large number of viewpoints need to be transmitted, to a point where the high bandwidth utilisation becomes a major concern. One approach to this problem is the utilisation of the so called “depth maps” in Depth Image Based Rendering (DIBR) techniques, where different views, can be reconstructed (rendered) at the receiver side when required, with a fraction of the bandwidth. Therefore the quality and accuracy of the information present in depth maps and its ability to reconstruct the required views has become the subject of much greater scrutiny in multimedia research. Several challenges arise when assessing the quality of depth maps, such as the lack of a proper reference for comparing the available depth maps, especially when it comes to live High Definition (HD) content. The first contribution of this thesis focuses on developing a novel subjective assessment approach, which addresses the presence of disocclusions in rendered views. The goal of this assessment approach is to enable the subjective evaluation of rendered views, to provide results that are more representative of the quality of the depth map utilised in the rendering process. The adopted approach performance has been evaluated through correlating the obtained subjective results with well-established objective metric measurements, such as PSNR, PSPNR, SSIM and VQM. The second contribution of this thesis is concerned with establishing a test data set, which includes different colour sequences together with various depth estimation algorithms and different depth post processing techniques. State of the art depth estimation algorithms were examined such as RSGM, DERS and HRM, in order to obtain a wide range of depth map qualities. The depth map data set is utilised in the DIBR process to generate rendered views, which are in turn subjectively assessed utilising the approach developed in contribution one of this thesis. The assessment is carried out in both a 2D and 3D setup and statistical analysis is utilised to establish observations and conclusions over the depth map performance. The third and final contribution of this thesis is related to a non-reference evaluation approach in assessing the quality of the depth maps utilised. This approach exploits the edges present in the available depth maps and compares that to the edges in the corresponding colour views. Edge pixels in depth maps are then classified into correct and error edge pixels. The obtained results are then correlated with the subjective results obtained in the second contribution. The result analysis of this non-reference model provides clear indications of depth map performance and its dependency on the associated colour sequence selection. The high correlation values with the subjective results were in the range of 75-81%. These correlation values are of more significance when compared to best performing quality metrics (e.g. VQM obtained correlation values of 59%) under the researched scenarios.

Item Type: Thesis (Doctoral)
Divisions : Theses
Authors :
Date : 29 January 2016
Funders : Self-funded
Contributors :
ContributionNameEmailORCID,, Ahmet
Depositing User : Nasser Haddad
Date Deposited : 09 Feb 2016 10:38
Last Modified : 31 Oct 2017 17:59

Actions (login required)

View Item View Item


Downloads per month over past year

Information about this web site

© The University of Surrey, Guildford, Surrey, GU2 7XH, United Kingdom.
+44 (0)1483 300800