Perceptual Depth Quality Assessment of Stereoscopic Omnidirectional Images

TL;DR

This paper introduces a no-reference Depth Quality Index (DQI) for assessing depth quality in stereoscopic omnidirectional images, motivated by human visual system cues. The method leverages interocular discrepancy across color channels, adaptive viewport sampling on the equator, and single-level Haar frequency decomposition, with statistics—standard deviation and entropy intensity—fed into an SVR to predict depth quality; it is extendable to depth-guided overall QoE by combining with existing IQA features. Experimental results on SOLID and Waterloo datasets show that DQI outperforms state-of-the-art IQA and DQA metrics for depth prediction, and that incorporating DQI improves overall QoE prediction. The work provides a practical, efficient framework for depth-aware quality assessment in 3D OI content, with potential applications in VR systems and streaming pipelines.

Abstract

Depth perception plays an essential role in the viewer experience for immersive virtual reality (VR) visual environments. However, previous research investigations in the depth quality of 3D/stereoscopic images are rather limited, and in particular, are largely lacking for 3D viewing of 360-degree omnidirectional content. In this work, we make one of the first attempts to develop an objective quality assessment model named depth quality index (DQI) for efficient no-reference (NR) depth quality assessment of stereoscopic omnidirectional images. Motivated by the perceptual characteristics of the human visual system (HVS), the proposed DQI is built upon multi-color-channel, adaptive viewport selection, and interocular discrepancy features. Experimental results demonstrate that the proposed method outperforms state-of-the-art image quality assessment (IQA) and depth quality assessment (DQA) approaches in predicting the perceptual depth quality when tested using both single-viewport and omnidirectional stereoscopic image databases. Furthermore, we demonstrate that combining the proposed depth quality model with existing IQA methods significantly boosts the performance in predicting the overall quality of 3D omnidirectional images.

Figures (8)

• Figure 1: An end-to-end 3D omnidirectional content processing pipeline.
• Figure 2: Examples of JPEG compressed 3D OIs with left and right ERP images. Both left and right views have the same compression level separately (level 1 versus level 4). Subjective ground-truth labels are provided. IQ: image quality, DQ: depth quality, OQ: overall QoE.
• Figure 3: Diagram of the proposed depth quality measure and the depth-guided overall QoE measure, where S represents subtraction operation.
• Figure 4: Interocular discrepancy maps with different depth levels. (a) Left view with medium disparity; (b) Right view with medium disparity; (c) Interocular discrepancy map of (a) and (b); (d) Left view with large disparity; (e) Right view with large disparity; (f) Interocular discrepancy map of (d) and (e).
• Figure 5: Demonstration of color decomposition and viewport selection for Fig. \ref{['fig4']} (f). (a-c) Luminance and two chroma components in LAB color space, respectively; (d-f) The corresponding viewports extracted from (a-c).