Objective Video Quality Assessment Methods: A Classification, Review, and Performance Comparison

Abstract

Abstract-With the increasing demand for video-based applications, the reliable prediction of video quality has increased in importance. Numerous video quality assessment methods and metrics have been proposed over the past years with varying computational complexity and accuracy. In this paper, we introduce a classification scheme for full-reference and reduced-reference media-layer objective video quality assessment methods. Our classification scheme first classifies a method according to whether natural visual characteristics or perceptual (human visual system) characteristics are considered. We further subclassify natural visual characteristics methods into methods based on natural visual statistics or natural visual features. We subclassify perceptual characteristics methods into frequency-or pixel-domain methods. According to our classification scheme, we comprehensively review and compare the media-layer objective video quality models for both standard resolution and high definition video. We find that the natural visual statistics based MultiScale-Structural SIMilarity index (MS-SSIM), the natural visual feature based Video Quality Metric (VQM), and the perceptual spatio-temporal frequency-domain based MOtion-based Video Integrity Evaluation (MOVIE) index give the best performance for the LIVE Video Quality Database.Index Terms-Full-reference metric, objective video quality, perceptual video quality, reduced-reference metric.

A novel methodology for mapping objective video quality metrics to the subjective MOS scale

“…The Peak Signal-to-Noise Ratio (PSNR), is the most widely used FR objective metric due to its simplicity, even though is often criticised for having a poor correlation with the subjective tests [16]. A multitude of more complex metrics based on natural visual characteristics, or that aim to model the Human Visual System (HVS), have also been proposed [4]. These metrics incorporate factors such as colour perception, contrast sensitivity or pattern masking, with one example being the Structural Similarity Index (SSIM) metric and its different variations [17].…”

“…The databases contain the MOS scores for each test sequence averaged across all participants as in (4).…”

“…Researchers and industry alike have recognised this, and much work has been conducted in order to develop objective VQA metrics that can be implemented with multimedia services for automatic estimation of video quality without the need for the users to provide their opinion [4].…”

Objective Video Quality Assessment — Towards Large Scale Video Database Enhanced Model Development

“…They are typically targeted towards a certain measurement situation such as near-lossless quality, or lowbitrate scenarios, or packet loss situations. Depending on whether the full reference video is available for measure- A recent overview of video quality algorithms including a sophisticated categorization can be found in [8]. In [9], the authors focus on identification of internal indicators used by some well-known algorithms such that they may be exploited independently to widen the scope of application.…”

Quality Assessment of User-Generated Video Using Camera Motion

“…(2) The commonly-used video QA measurements such as signal-to-noise ratio (SNR), peak-signal-to-noise ratio (PSNR) and mean squared error (MSE) [5], are computationally simply, however, they disregard the characteristics of human visual perception. (3) A lot of research interest has been focused on objective video QA [2,6], however, methods to assess the visual quality of digital video as perceived by human observer are becoming increasingly important, due to the large number of applications that target humans as the end users of video. The only reliable method to assess the video quality is to ask human subjects for their own opinions, which is termed subjective video QA.…”