Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Subjective and Objective Quality Assessment of Non-Uniformly Distorted Omnidirectional Images

Jiebin Yan, Jiale Rao, Xuelin Liu, Yuming Fang, Yifan Zuo, Weide Liu

TL;DR

This work tackles non-uniform distortions in omnidirectional image quality assessment by introducing JUFE-10K, a large-scale subjective database of 10,320 non-uniformly distorted OIs and accompanying viewing-behavior data. It proposes OIQAND, a perception-guided OIQA model that uses a Swin-transformer-based viewport backbone, a distortion-adaptive perception module, and multi-scale feature fusion to predict viewport-wise quality and overall QoE. Empirical results show OIQAND outperforms state-of-the-art 2D-IQA and OIQA methods on JUFE-10K, demonstrating the importance of modeling non-uniform distortion and viewer behavior. The work advances practical VR QoE assessment and provides a resource for future research in non-uniform distortion-aware OIQA and scanpath-aware modeling.

Abstract

Omnidirectional image quality assessment (OIQA) has been one of the hot topics in IQA with the continuous development of VR techniques, and achieved much success in the past few years. However, most studies devote themselves to the uniform distortion issue, i.e., all regions of an omnidirectional image are perturbed by the ``same amount'' of noise, while ignoring the non-uniform distortion issue, i.e., partial regions undergo ``different amount'' of perturbation with the other regions in the same omnidirectional image. Additionally, nearly all OIQA models are verified on the platforms containing a limited number of samples, which largely increases the over-fitting risk and therefore impedes the development of OIQA. To alleviate these issues, we elaborately explore this topic from both subjective and objective perspectives. Specifically, we construct a large OIQA database containing 10,320 non-uniformly distorted omnidirectional images, each of which is generated by considering quality impairments on one or two camera len(s). Then we meticulously conduct psychophysical experiments and delve into the influence of both holistic and individual factors (i.e., distortion range and viewing condition) on omnidirectional image quality. Furthermore, we propose a perception-guided OIQA model for non-uniform distortion by adaptively simulating users' viewing behavior. Experimental results demonstrate that the proposed model outperforms state-of-the-art methods. The source code is available at https://github.com/RJL2000/OIQAND.

Subjective and Objective Quality Assessment of Non-Uniformly Distorted Omnidirectional Images

TL;DR

This work tackles non-uniform distortions in omnidirectional image quality assessment by introducing JUFE-10K, a large-scale subjective database of 10,320 non-uniformly distorted OIs and accompanying viewing-behavior data. It proposes OIQAND, a perception-guided OIQA model that uses a Swin-transformer-based viewport backbone, a distortion-adaptive perception module, and multi-scale feature fusion to predict viewport-wise quality and overall QoE. Empirical results show OIQAND outperforms state-of-the-art 2D-IQA and OIQA methods on JUFE-10K, demonstrating the importance of modeling non-uniform distortion and viewer behavior. The work advances practical VR QoE assessment and provides a resource for future research in non-uniform distortion-aware OIQA and scanpath-aware modeling.

Abstract

Omnidirectional image quality assessment (OIQA) has been one of the hot topics in IQA with the continuous development of VR techniques, and achieved much success in the past few years. However, most studies devote themselves to the uniform distortion issue, i.e., all regions of an omnidirectional image are perturbed by the ``same amount'' of noise, while ignoring the non-uniform distortion issue, i.e., partial regions undergo ``different amount'' of perturbation with the other regions in the same omnidirectional image. Additionally, nearly all OIQA models are verified on the platforms containing a limited number of samples, which largely increases the over-fitting risk and therefore impedes the development of OIQA. To alleviate these issues, we elaborately explore this topic from both subjective and objective perspectives. Specifically, we construct a large OIQA database containing 10,320 non-uniformly distorted omnidirectional images, each of which is generated by considering quality impairments on one or two camera len(s). Then we meticulously conduct psychophysical experiments and delve into the influence of both holistic and individual factors (i.e., distortion range and viewing condition) on omnidirectional image quality. Furthermore, we propose a perception-guided OIQA model for non-uniform distortion by adaptively simulating users' viewing behavior. Experimental results demonstrate that the proposed model outperforms state-of-the-art methods. The source code is available at https://github.com/RJL2000/OIQAND.
Paper Structure (32 sections, 14 equations, 7 figures, 5 tables)

This paper contains 32 sections, 14 equations, 7 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Subjective OIQA Databases
  4. Objective 2D-IQA Models
  5. Objective OIQA Models
  6. Subjective Experiment Methodology
  7. Database Construction
  8. Psychophysical Experiment Design
  9. Viewing Conditions
  10. Subjective Methodology
  11. Subjective Data Processing and Analysis
  12. Data Processing
  13. Data Analysis
  14. The Proposed OIQA Model
  15. Viewport Extraction
  16. ...and 17 more sections

Figures (7)

  • Figure 1: A visual example of how users explore omnidirectional images.
  • Figure 2: Thumbnails of reference OIs in the proposed JUFE-10K database.
  • Figure 3: The scatter diagram of spatial information (SI) and colorfulness (CF) of six OIQA databases.
  • Figure 4: Visual examples of non-uniform distortion. Zoom-in for better view.
  • Figure 5: The statistic distributions of the quality of OIs from different perspectives.
  • ...and 2 more figures