Cognitive no-reference video quality assessment for mobile streaming services

IEEE Signal Process. Lett.

Famaey

et al. 2017

Self Cite

Abstract-Video content providers put stringent requirements on the quality assessment methods realized on their services. They need to be accurate, real-time, adaptable to new content, and scalable as the video set grows. In this letter, we introduce a novel automated and computationally efficient video assessment method. It enables accurate real-time (online) analysis of delivered quality in an adaptable and scalable manner. Offline deep unsupervised learning processes are employed at the server side and inexpensive no-reference measurements at the client side. This provides both real-time assessment, as well as performance comparable to the full reference counter-part, while maintaining its no-reference characteristics. We tested our approach on the LIMP Video Quality Database (an extensive packet loss impaired video-set) obtaining a correlation between 78% and 91% to the FR benchmark (the Video Quality Metric, VQM). Due to its unsupervised learning essence, our method is flexible, dynamically adaptable to new content and scalable with the number of videos.Index Terms-Deep learning, unsupervised learning, video quality assessment, multimedia video services.

Section: Real-time Cognitive Video Quality Assessment Methodsmentioning

confidence: 99%

Deep Learning for Quality Assessment in Live Video Streaming

IEEE Signal Process. Lett.

Famaey

et al. 2017

Self Cite

“…In [61] we showed the use of Reinforcement Learning to optimize video quality in adaptive streaming, without 600 using complex heuristics. In [1] we showed how artificial neural networks could determine a linear combination of blur and noise that performed significantly better than these two NR metrics in isolation. Finally, our recent survey of machine learning in NR video quality assessment [62] provides a snapshot of the state-of-the-art on which our work is based.…”

Section: Related Work On Machine Learning For Nr Quality Assessmentmentioning

confidence: 99%

“…Low complexity No-Reference (NR) video quality methods have the potential to provide real-time video quality assessment and automated quality control, for instance in the context of video streaming on demand [1], peer to peer services [2,3] or real-time network management [4,5]. This is because simple Due to their particular methodology, computational requirements and functional limitations, neither FR methods nor subjective evaluations are viable 10 to automate quality control processes, whereby both scalability and speed are required.…”

Section: Introductionmentioning

confidence: 99%

“…Each sample was then impaired at twelve packet loss rates, obtaining 960 videos as specified in Table 1. 1 Next, we carried out a detailed evaluation of the whole dataset, according to all the NR metrics under scrutiny, including also the FR metric SSIM. Blur, 145 noise, blockiness, ringing or temporal impairments have been quantified for measuring the end-user's quality [15].…”

mentioning

confidence: 99%

See 1 more Smart Citation

Predictive no-reference assessment of video quality

Signal Processing: Image Communication

Stavrou

et al. 2017

Self Cite

Among the various means to evaluate the quality of video streams, light-weight No-Reference (NR) methods have low computation and may be executed on thin clients. Thus, these methods would be perfect candidates in cases of real-time quality assessment, automated quality control and in adaptive mobile streaming. Yet, existing real-time, NR approaches are not typically designed to tackle network distorted streams, thus performing poorly when compared to Full-Reference (FR) algorithms. In this work, we present a generic NR method whereby machine learning (ML) may be used to construct a quality metric trained on simplistic NR metrics. Testing our method on nine, representative ML algorithms allows us to show the generality of our approach, whilst finding the best-performing algorithms. We use an extensive video dataset (960 video samples), generated under a variety of lossy network conditions, thus verifying that our NR metric remains accurate under realistic streaming scenarios. In this way, we achieve a quality index that is comparably as computationally efficient as typical NR metrics and as accurate as the FR algorithm Video Quality Metric (97% correlation).

“…Pandremmenou et al [24] employed the Least Absolute Shrinkage and Selection Operator (LASSO) regression method for assessing the accuracy of bitstream parameters to FR metrics and subjective analysis in videos affected by compression and synthetic impairments. In our previous research we developed a lightweight algorithm combining bitstream parameters (video bitrate, complexity and motion) with pixel artifacts (blur and noise) [36]. We presented the machine learning-based algorithm showing high correlation with SSIM.…”

Section: Introductionmentioning

confidence: 99%

A Regression Method for real-time video quality evaluation

Proceedings of the 14th International Conference on Advances in Mobile Computing and Multi Media

Liotta

2016

Self Cite

No-Reference (NR) metrics provide a mechanism to assess video quality in an ever-growing wireless network. Their low computational complexity and functional characteristics make them the primary choice when it comes to realtime content management and mobile streaming control. Unfortunately, common NR metrics suffer from poor accuracy, particularly in network-impaired video streams. In this work, we introduce a regression-based video quality metric that is simple enough for real-time computation on thin clients, and comparably as accurate as state-of-the-art FullReference (FR) metrics, which are functionally and computationally inviable in real-time streaming. We benchmark our metric against the FR metric VQM (Video Quality Metric), finding a very strong correlation factor.