Collaborative Edge-Assisted Systems for HTTP Adaptive Video Streaming

5G/6G Innovation Center,  University of Surrey, UK

6th January 2023 | Guildford, UK

Abstract: The proliferation of novel video streaming technologies, advancement of networking paradigms, and steadily increasing numbers of users who prefer to watch video content over the Internet rather than using classical TV have made video the predominant traffic on the Internet. However, designing cost-effective, scalable, and flexible architectures that support low-latency and high-quality video streaming is still a challenge for both over-the-top (OTT) and ISP companies. In this talk, we first introduce the principles of video streaming and the existing challenges. We then review several 5G/6G networking paradigms and explain how we can leverage networking technologies to form collaborative network-assisted video streaming systems for improving users’ quality of experience (QoE) and network utilization.

Reza Farahani is a last-year Ph.D. candidate at the University of Klagenfurt, Austria, and a Ph.D. visitor at the University of Surrey, Uk. He received his B.Sc. in 2014 and M.Sc. in 2019 from the university of Isfahan, IRAN, and the university of Tehran, IRAN, respectively. Currently, he is working on the ATHENA project in cooperation with its industry partner Bitmovin. His research is focused on designing modern network-assisted video streaming solutions (via SDN, NFV, MEC, SFC, and P2P paradigms), multimedia Communication, computing continuum challenges, and parallel and distributed systems. He also worked in different roles in the computer networks field, e.g., network administrator, ISP customer support engineer, Cisco network engineer, network protocol designer, network programmer, and Cisco instructor (R&S, SP).

IEEE/IFIP Network Operations and Management Symposium (NOMS)

8-12 May 2023 | Miami, FL, USA

Conference website

[PDF]

Farzad Tashtarian (AAU, Austria), Abdelhak Bentaleb (Concordia University, Canada), Hadi Amirpour (AAU, Austria), Babak Taraghi (AAU, Austria), Christian Timmerer (AAU, Austria), Hermann Hellwagner (AAU, Austria), Roger Zimmermann (National University of Singapore, Singapore)

Video content in Live HTTP Adaptive Streaming (HAS) is typically encoded using a pre-defined, fixed set of bitrate-resolution pairs (termed Bitrate Ladder), allowing playback devices to adapt to changing network conditions using an adaptive bitrate (ABR) algorithm. However, using a fixed one-size-fits-all solution when faced with various content complexities, heterogeneous network conditions, viewer device resolutions and locations, does not result in an overall maximal viewer quality of experience (QoE). Here, we consider these factors and design LALISA, an efficient framework for dynamic bitrate ladder optimization in live HAS. LALISA dynamically changes a live video session’s bitrate ladder, allowing improvements in viewer QoE and savings in encoding, storage, and bandwidth costs. LALISA is independent of ABR algorithms and codecs, and is deployed along the path between viewers and the origin server. In particular, it leverages the latest developments in video analytics to collect statistics from video players, content delivery networks and video encoders, to perform bitrate adder tuning. We evaluate the performance of LALISA against existing solutions in various video streaming scenarios using a trace-driven testbed. Evaluation results demonstrate significant improvements in encoding computation (24.4%) and bandwidth (18.2%) costs with an acceptable QoE.

CD-LwTE: Cost- and Delay-aware Light-weight Transcoding at the Edge

IEEE Transactions on Network and Service Management (TNSM)

[PDF]

Alireza Erfanian (Alpen-Adria-Universität Klagenfurt, Austria), Hadi Amirpour (Alpen-Adria-Universität Klagenfurt, Austria), Farzad Tashtarian (Alpen-Adria-Universität Klagenfurt, Austria), Christian Timmerer (Alpen-Adria-Universität Klagenfurt, Austria), and Hermann Hellwagner.

Abstract—The edge computing paradigm brings cloud capabilities close to the clients. Leveraging the edge’s capabilities can improve video streaming services by employing the storage capacity and processing power at the edge for caching and transcoding tasks, respectively, resulting in video streaming services with higher quality and lower latency. In this paper, we propose CD-LwTE, a Cost- and Delay-aware Light-weight Transcoding approach at the Edge, in the context of HTTP Adaptive Streaming (HAS). The encoding of a video segment requires computationally intensive search processes. The main idea of CD-LwTE is to store the optimal search results as metadata for each bitrate of video segments and reuse it at the edge servers to reduce the required time and computational resources for transcoding. Aiming at minimizing the cost and delay of Video-on-Demand (VoD) services, we formulate the problem of selecting an optimal policy for serving segment requests at the edge server, including (i) storing at the edge server, (ii) transcoding from a higher bitrate at the edge server, and (iii) fetching from the origin or a CDN server, as a Binary Linear Programming (BLP) model. As a result, CD-LwTE stores the popular video segments at the edge and serves the unpopular ones by transcoding using metadata or fetching from the origin/CDN server. In this way, in addition to the significant reduction in bandwidth and storage costs, the transcoding time of a requested segment is remarkably decreased by utilizing its corresponding metadata. Moreover, we prove the proposed BLP model is an NP-hard problem and propose two heuristic algorithms to mitigate the time complexity of CD-LwTE. We investigate the performance of CD-LwTE in comprehensive scenarios with various video contents, encoding software, encoding settings, and available resources at the edge. The experimental results show that our approach (i) reduces the transcoding time by up to 97%, (ii) decreases the streaming cost, including storage, computation, and bandwidth costs, by up to 75%, and (iii) reduces delay by up to 48% compared to state-of-the-art approaches.

IEEE Transactions on Circuits and Systems for Video Technology (IEEE TCSVT)

[PDF]

Hadi Amirpour (Alpen-Adria-Universität Klagenfurt, Austria), Mohammad Ghanbari (University of Essex, UK), and Christian Timmerer (Alpen-Adria-Universität Klagenfurt, Austria)

Abstract: In HTTP Adaptive Streaming (HAS), each video is divided into smaller segments, and each segment is encoded at multiple pre-defined bitrates to construct a bitrate ladder. To optimize bitrate ladders, per-title encoding approaches encode each segment at various bitrates and resolutions to determine the convex hull. From the convex hull, an optimized bitrate ladder is constructed, resulting in an increased Quality of Experience (QoE) for end-users. With the ever-increasing efficiency of deep learning-based video enhancement approaches, they are more and more employed at the client-side to increase the QoE, specifically when GPU capabilities are available. Therefore, scalable approaches are needed to support end-user devices with both CPU and GPU capabilities (denoted as CPU-only and GPU-available end-users, respectively) as a new dimension of a bitrate ladder.
To address this need, we propose DeepStream, a scalable content-aware per-title encoding approach to support both CPU-only and GPU-available end-users. (i) To support backward compatibility, DeepStream constructs a bitrate ladder based on any existing per-title encoding approach. Therefore, the video content will be provided for legacy end-user devices with CPU-only capabilities as a base layer (BL). (ii) For high-end end-user devices with GPU capabilities, an enhancement layer (EL) is added on top of the base layer comprising lightweight video super-resolution deep neural networks (DNNs) for each bitrate-resolution pair of the bitrate ladder. A content-aware video super-resolution approach leads to higher video quality, however, at the cost of bitrate overhead. To reduce the bitrate overhead for streaming content-aware video super-resolution DNNs, DeepCABAC, context-adaptive binary arithmetic coding for DNN compression, is used. Furthermore, the similarity among (i) segments within a scene and (ii) frames within a segment are used to reduce the training costs of DNNs.
Experimental results show bitrate savings of 34% and 36% to maintain the same PSNR and VMAF, respectively, for GPU-available end-users, while the CPU-only users get the desired video content as usual.

Keywords—HTTP adaptive streaming, per-title encoding, video streaming, video super-resolution.

Elsevier Signal Processing

[PDF]

Ningxiong Mao (Southwest Jiaotong University), Hongjie Hea (Southwest Jiaotong University), Fan Chenb (Southwest Jiaotong University), Lingfeng Qu (Southwest Jiaotong University), Hadi Amirpour (Alpen-Adria-Universität Klagenfurt, Austria), and Christian Timmerer (Alpen-Adria-Universität Klagenfurt, Austria)

Abstract:

Color image Reversible Data Hiding (RDH) is getting more and more important since the number of its applications is steadily growing. This paper proposes an efficient color image RDH scheme based on pixel value ordering (PVO), in which the channel correlation is fully utilized to improve the embedding performance. In the proposed method, the channel correlation is used in the overall process of data embedding, including prediction stage, block selection and capacity allocation. In the prediction stage, since the pixel values in the co-located blocks in different channels are monotonically consistent, the large pixel values are collected preferentially by pre-sorting the intra-block pixels. This can effectively improve the embedding capacity of RDH based on PVO. In the block selection stage, the description accuracy of block complexity value is improved by exploiting the texture similarity between the channels. The smoothing the block is then preferentially used to reduce invalid shifts. To achieve low complexity and high accuracy in capacity allocation, the proportion of the expanded prediction error to the total expanded prediction error in each channel is calculated during the capacity allocation process. The experimental results show that the proposed scheme achieves significant superiority in fidelity over a series of state-of-the-art schemes. For example, the PSNR of the Lena image reaches 62.43dB, which is a 0.16dB gain compared to the best results in the literature with a 20,000bits embedding capacity.

Keywords—Reversible data hiding, color image, pixel value ordering, channel correlation