On the Impact of Viewing Distance on Perceived Video Quality
Hadi Amirpour 1
Raimund Schatz 2
Christian Timmerer 1
Mohammad Ghanbari 1,3
1
Christian Doppler Laboratory ATHENA, Alpen-Adria-Universität, Klagenfurt, Austria
2
AIT Austrian Institute of Technology, Austria 3
School of Computer Science and Electronic Engineering, University of Essex, UK
Paper ID: 093
VCIP 2021
Introduction
Due to the growing importance of optimizing the quality and efficiency of video streaming
delivery, accurate assessment of user-perceived video quality becomes increasingly impor-
tant.
Due to the wide range of viewing distances encountered in real-world viewing settings, the
perceived video quality can vary significantly in everyday viewing situations.
Figure 1. There is a wide range of viewing distances in real-world viewing settings.
Although some studies in this direction already exist, their focus mainlywas on (low resolution)
images rather than on video sequences.
We argue that in the context of video quality the relation between the viewing distance and
the perceived quality should be more systematically investigated.
In this paper, we empirically study the impact of viewing distance on perceived video quality.
Additionally, the impact of viewing distance on the correlation between common objective
metrics and subjective scores is investigated to answer the following two research questions:
RQ1: How does varying the viewing distance influence subjective video quality ratings?
RQ2: How does changing the viewing distance alter the correlation between the estimated
objective video quality metrics and subjective ground truth data?
Subjective Experiment Design
Test Content
Based on SI and TI, video sequences were grouped into five clusters using k-means clustering.
In total, 10 sequences (two from each cluster) have been selected for the subjective test.
All sequences were 8-bit, 1080p in resolution, and had 5 or 10 seconds duration.
x265 (version 3.4) was used to encode video sequences at three quality/distortion levels,
namely excellent (imperceptible), good (perceptible, but not annoying), and fair quality (slightly
annoying) determined by an expert review.
Test Environment
Our laboratory was set up according to ITU BT.500-13.
We used a Samsung LE46C750R2Z 46” Full HD TV as display, color calibrated, and with all
image enhancements turned off.
We calculated the (closest) optimal viewing distance using the following equation:
V D =
DS
q
(NHR
NV R)2 + 1 × CV R × tan( 1
60)
(1)
where DS is display size, and NHR and NVR are the display’s native horizontal and vertical
resolutions, respectively. CVR represents the vertical resolution of the video.
For our experimental setup, the optimal viewing distance d1 was equal to 3.2H = 1.82m, with
H being the height of the display.
We added two non-optimal, larger viewing distances d2, d3 to our test design, resulting in the
following three distance levels:
d1= 1.82m (3.2H), d2= 2.80m (4.9H), d3= 3.90m (6.85H)
The latter two additional distances were chosen because they represent more common view-
ing distances used in practice.
Test Protocol
Absolute Category Rating (ACR) was selected as testing method.
Participants rated the overall perceived quality of the video sequences using a 7-point Likert
scale. Each subject scored 90 stimuli in total, which took 23 minutes per subject on average.
In total, 25 subjects (avg. age 28; 14 male and 11 female) participated in the subjective test.
Video sequences and the order of distances were randomly selected for each subject.
Evaluation Results
The impact of the viewing distance on the perceived video quality (MOS) at different encoding
quality levels per each clip (RQ1) is shown in Fig. 2.