Quality Metric Aggregation for HDR/WCG Images

Abstract

High Dynamic Range (HDR) and Wide Color Gamut (WCG) screens are able to display images with brighter and darker pixels with more vivid colors than ever. Automatically assessing the quality of these HDR/WCG images is of critical importance to evaluate the performances of image compression schemes. In recent years, full-reference metrics, such as HDR-VDP-2, PU-encoding metrics, have been designed for this purpose. However, none of these metrics consider chromatic artifacts. In this paper, we propose our own full-reference quality metric adapted to HDR and WCG content that is sensitive to chromatic distortions. The proposed metric is based on two existing HDR quality metrics and color image features. A support vector machine regression is used to combine the aforementioned features. Experimental results demonstrate the effectiveness of the proposed metric in the context of image compression.

Image quality evaluation for high dynamic range and wide color gamut applications using visual spatial processing of color differences

“…• We present the best color difference quality metric that has a certain level of simplicity (eg, no filter banks). As such, we will not evaluate the currently best performing image quality metrics for HDR imaging such as HDR-VDP-2 35,36 or HDR-VQM, 37 or newer approaches combining such metrics using machine learning techniques [38][39][40][41][42] which are significantly more computationally complex than the metrics described in this article. • Since we have previously shown that the per-pixel ΔE ITP metric outperforms other per-pixel metrics 1 , in this article, we motivate and present two spatial extensions to the ΔE ITP metric that produce state-of-the-art results when compared to per-pixel color difference metrics.…”

“… Given the current trends toward widespread adoption of HDR and WCG, we provide insight into the differences between HDR and SDR and motivate why developing a metric for HDR and WCG applications is important. We present the best color difference quality metric that has a certain level of simplicity (eg, no filter banks). As such, we will not evaluate the currently best performing image quality metrics for HDR imaging such as HDR‐VDP‐2 35,36 or HDR‐VQM, 37 or newer approaches combining such metrics using machine learning techniques 38‐42 which are significantly more computationally complex than the metrics described in this article. Since we have previously shown that the per‐pixel Δ E ITP metric outperforms other per‐pixel metrics 1 , in this article, we motivate and present two spatial extensions to the Δ E ITP metric that produce state‐of‐the‐art results when compared to per‐pixel color difference metrics. We presented preliminary versions of the spatial extension of Δ E ITP in our previous work 2,3 .…”

A Viewport-Driven Multi-Metric Fusion Approach for 360-Degree Video Quality Assessment

“…On the other hand, the use of multiple objec-tive metrics computed on these viewports allow our method to have a good performance for the complex and diversifying nature of visual distortions appearing in 360-degree videos. Indeed, previous work in both traditional 2D [8,9,10] and 360-degree [7] VQA have recognized that even with the multitude of available objective IQA metrics, there is no single one that always performs best for all distortions. The combination of multiple metrics is thus a promising approach that can take advantages of the power of individual metrics to correlate with subjective scores on different distortions [11,10].…”

“…Indeed, previous work in both traditional 2D [8,9,10] and 360-degree [7] VQA have recognized that even with the multitude of available objective IQA metrics, there is no single one that always performs best for all distortions. The combination of multiple metrics is thus a promising approach that can take advantages of the power of individual metrics to correlate with subjective scores on different distortions [11,10]. Experimental results, based on the largest publicly available 360-degree video quality dataset, VQA-ODV [12], show the viability of our proposal, which outperforms state-of-the-art methods for 360-degree quality assessment.…”

Multi-Feature 360 Video Quality Estimation

“…On the other hand, the use of multiple objective metrics computed on these viewports allows our method to have a good performance for the complex and diverse nature of visual distortions appearing in 360-degree videos. Indeed, previous work in both traditional 2D [25], [32], [34] and 360-degree [3] VQA have recognized that even with the multitude of available objective IQA metrics, there is no single one that always performs best for all 1. Compared to [4], we extend the individual features used by our model, propose an adapted temporal pooling method, and provide an extensive new set of experiments, including different regression methods and a crossdataset validation.…”