Forecasting of viewers’ discomfort
- Author: Anastasiia Antsiferova
- Supervisor: dr. Dmitriy Vatolin
Introduction
Nowadays, numerous movies are produced in stereoscopic format. Despite the development of stereoscopic movie production, stereoscopic artifacts causing discomfort right up to headaches continue to appear even in high-budget movies.
In this study, the influence of geometric, color and temporal artifacts was examined.
Experiments
In a series of experiments, participants were asked to evaluate the level of discomfort while watching a specially prepared stereoscopic video. Over 300 people took part.
Experiment comparison
University | #subjects | #subjects per video | Duration | #videos | #videos per test | #scores | Year | 3D tech. | Stimuli |
---|---|---|---|---|---|---|---|---|---|
University of Surrey | 30 | 30 | 25 | 40 | 40 | 1200 | 2010 | Autostereo. | Synthetic S3D seq., encoding with different QP |
Catholic University of Korea | 20 | 20 | 18 | 36 | 36 | 720 | 2011 | Passive | Captured S3D seq., different parallax and motion |
Telecom Innovation Labs | 24 | 24 | 50 | 64 | 64 | 1536 | 2011 | Active | Open S3D DB, different parallax |
Philips Research Labs | 24 | 24 | 24 | 7 | 7 | 168 | 2011 | Autostereo. | 3D movie (converted), different parallax |
Beijing Institute of Ophthal. | 30 | 30 | 30 | 1 | 1 | 30 | 2012 | Active+passive | 3D movie (captured) |
LUNAM University | 29 | 29 | 28 | 110 | 110 | 3190 | 2012 | Active | Open S3D DB (NAMA3DS1-COSPAD1), different degradations |
Yonsei University | 28 | 28 | 29 | 110 | 110 | 3080 | 2013 | Passive | Open S3D DB,10 degradation types |
Acreo Institute | 48 | 28 | 29 | 110 | 110 | 5280 | 2013 | Passive | Open S3D DB,10 degradation types |
Tampere University of Tech. | 10 | 10 | 45 | 40 | 40 | 400 | 2013 | Passive | Captured S3D seq., encoding with different QP |
Roma Tre University | 854 | 43-255 | 90 | - | 1 | 854 | 2014 | Passive | Screening after watching S3D movies in cinemas |
Yonsei University | 56 | 56 | 10.5 | 30.1 | 31 | 1736 | 2014 | Autostereo. | Open S3D DB, different SI and TI |
University College Dublin | 40 | 4 | 5.5 | 25.11 | 33 | 1320 | 2015 | Active | Open S3D DB, packet losses |
University of British Columbia | 88 | 88 | 48 | 208 | 208 | 18304 | 2015 | Passive | Open S3D DB, different SI/TI and compression degradations |
University North | 146 | 18 | 8 | 184 | 26 | 146 | 2016 | Active | Open S3D DB, 22 degradation types (packet losses, encoding, resizing, disparity, brightness, geometry etc.) |
University of Coimbra | 35 | 2-6 | 7-1 | 184 | 26 | 146 | 2016 | Active | Open S3D DB, 22 degradation types (packet losses, encoding, resizing, disparity, brightness, geometry etc.) |
Lomonosov MSU | 302 | 370 | 40 | 60 | 60 | 22200 | 2017 | Passive | Scenes from 3D movies (captured), 4 types of S3D distortions, 5 intensities |
Experiment process
Discomfort level sorted by mean discomfort (red for high)
Here “C” is the color mismatch between stereoscopic views, “R” is the rotation mismatch, “T” is the temporal shift and “S” is the scale mismatch. The distortion intensity is indicated by numbers from 0 to 4, where 0 corresponds to the absence of distortion (the original scene was demonstrated).
To solve the problem of assessing the discomfort level, we used machine learning algorithms on the processed experimental data. More than 30,000 configurations were tested on the obtained stratified cross-validation dataset.
The best result was shown by a linear regression model with the Huber loss function and L2–regularization.
Results
The proposed models were applied to evaluate 60 stereoscopic movies.
The predicted level of discomfort that viewers may experience and the predicted percentage of viewers who will probably feel the discomfort while watching the analyzed movies are illustrated below:
Dataset
Dataset consists of 60 video fragments that have been viewed and evaluated by a group of 302 people
p-13 folder consists of:
- File with viewing order (direct and backward)
- 60 subfolders, each contains video fragment, it's grade and information about distortions in the video
Downloads
IC3D Paper (2017)
Accepted version of the paper: Download
Dataset: Download
Reference
Citation A. Antsiferova, D. Vatolin. “The influence of 3D video artifacts on discomfort of 302 viewers”. 2017 IEEE International Conference on 3D Immersion (IC3D). pp. 1-8.
Bibtex
@inproceedings{antsiferova2017influence,
title={The influence of 3D video artifacts on discomfort of 302 viewers},
author={Antsiferova, Anastasia and Vatolin, D},
booktitle={2017 International Conference on 3D Immersion (IC3D)},
pages={1--8},
year={2017},
organization={IEEE}
}
-
MSU Benchmark Collection
- Video Colorization Benchmark
- Super-Resolution for Video Compression Benchmark
- Defenses for Image Quality Metrics Benchmark
- Learning-Based Image Compression Benchmark
- Super-Resolution Quality Metrics Benchmark
- Video Saliency Prediction Benchmark
- Metrics Robustness Benchmark
- Video Upscalers Benchmark
- Video Deblurring Benchmark
- Video Frame Interpolation Benchmark
- HDR Video Reconstruction Benchmark
- No-Reference Video Quality Metrics Benchmark
- Full-Reference Video Quality Metrics Benchmark
- Video Alignment and Retrieval Benchmark
- Mobile Video Codecs Benchmark
- Video Super-Resolution Benchmark
- Shot Boundary Detection Benchmark
- The VideoMatting Project
- Video Completion
- Codecs Comparisons & Optimization
- VQMT
- MSU Datasets Collection
- Metrics Research
- Video Quality Measurement Tool 3D
- Video Filters
- Other Projects