Impacts of Retina-Related Zones on Quality Perception of Omnidirectional Image

Abstract

Virtual Reality (VR), which brings immersive experiences to viewers, has been gaining popularity in recent years. A key feature in VR systems is the use of omnidirectional content, which provides 360-degree views of scenes. In this work, we study the human quality perception of omnidirectional images, focusing on different zones surrounding the foveation point. For that purpose, an extensive subjective experiment is carried out to assess the perceptual quality of omnidirectional images with non-uniform quality. Through experimental results, the impacts of different zones are analyzed. Moreover, nineteen objective quality metrics, including foveal quality metrics, are evaluated using our database. It is quantitatively shown that the zones corresponding to the fovea and parafovea of human eyes are extremely important for quality perception, while the impacts of the other zones corresponding to the perifovea and periphery are small. Besides, the investigated metrics are found to be not effective enough to reflect the quality perceived by viewers.

An Effective Foveated 360° Image Assessment Based on Graph Convolution Network

“…In the literature, there have been a lot of studies on foveated contents (i.e., images or videos) [3], [7], [8], [9], [10], [11], [12], [6], and [13]. Among them, there are, however, only some on 360°contents [3], [10], [11], [12], and [6].…”

“…Many techniques have been proposed for evaluating image quality. Traditional technical image quality metrics that are leveraged commonly in FR-IQA and RR-IQA include Mean Squared Error (MSE), Frequency Mean Square Error (FMSE) [15], Universal Quality Index (UQI) [16], Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) [17], Foveated Peak Signal-to-Noise Ratio(FPSNR) [8], Multi-scale SSIM (MS-SSIM) [18], Information content Weighted SSIM (IW-SSIM), Noise Quality Measure (NQM) [19], Visual Information Fidelity (VIF) [20] Visual Information Fidelity in the pixel domain (VIFp) [21], Weight Signal-to-Noise Ratio (WSNR), Feature similarity index measure (FSIM), Feature similarity measure (FSIMc) for color image, Foveal feature similarity measure (F-SSIM) [22], Perceptual Similarity (PSIM) [23], Analysis of Distortion Distribution-based (ADD-SSIM) [24], Foveal Structural Similarity [25], and Foveated Wavelet image Quality Index (FWQI) [7], Generic Statistical Information Model (GSIM), Riesz Transforms based Feature Similarity (RF-SIM) [26], Information content Weighted PSNR (IW-PSNR) [27], Blind/Referenceless Image Spatial Quality Evaluator (BRISQUE) [28], No-reference Free Energy-Based Robust Metric (NFERM) [29], Spectral Residual based Similarity (SR-SIM) [30], and Weighted Viewport PSNR (W-VPSNR) [3]. Therefore in this paper, we will evaluate our solution's performance in terms of 25 metrics: MSE, FMSE, UQI, PSNR, FPSNR, SSIM, MS-SSIM, IW-SSIM, NQM, VIF, VIFp, WSNR, FSIM, FSIMc, F-SSIM, PSIM, ADD-SSIM, FWQI, GSIM, RFSIM, IW-PSNR, BRISQUE, NFERM, SR-SIM, and W-PSNR in order to have an insight into our proposed solution from variety of angles.…”

“…In the next subsections, the two main blocks and parameter settings of the FoVGCN model will be described in detail. According to human retina characteristics, the quality of omnidirectional images is divided into five regions (Fovea, Parafovea, Perifovea, Near periphery, Far periphery) [3]. It means that the perceived quality of each pixel in a distorted image changes from the center region to the outer regions.…”

“…As a second way to investigate, the Gaussian distribution is leveraged to describe the degradation of human attraction from the center to the outer region. This idea follows the concept of the human retina [3] in which user perception tends to be affected by the quality ofovea and parafovea zones from the center to the outer zone of an image. Because human eyes concentrate significantly on the fovea and parafovea zones of the human retina which is a small region in the viewport [3].…”

An Effective Foveated 360° Image Assessment Based on Graph Convolution Network

“…By utilizing encoding parameters (i.e., quantization parameters and resolutions) that had been recorded, the authors proposed a rendering solution that is indicated to be able to significantly improve rendering throughput about 10× without perceptual loss, in comparison to the traditional solution of uniform quality. [3] is the first study that could quantify the impacts of different retina regions on user quality perception. In particular, the authors performed a subjective quality assessment of foveated 360°images.…”