Back to search

Institutes outside Greece

On quality aware adaptation of internet video

Abstract

dc:description

The main issue in video streaming over the Internet is that of adaptation: due to the Internet’s “best-effort” service model, the availability of sufficient bandwidth to guarantee good quality is not always feasible. Instead, congestion control over shared resources is required, which leads to fluctuations in available bandwidth. These transmission rate variations introduce a first level of quality degradation. Moreover, the time-varying complexity of the underlying visual scenes requires significantly varying transmission bandwidth in order to achieve good quality; otherwise, fluctuations in quality are observed. To address these problems, many rate adaptation techniques have been proposed. However, these approaches either consider adaptation exclusively from a network perspective, such as TCP-friendliness, completely ignoring the impact on quality, or they use simplified quality metrics, such as peak signal-to-noise ratio (PSNR), which do not necessarily provide accurate representations of the quality as perceived by the video service user. This dissertation advocates the incorporation of emerging objective video quality metrics into the adaptation loop of Internet video. As a result of recent research efforts, objective quality metrics constitute computational models that produce quality assessments highly correlated with human judgments of quality. Taking into account the time-varying relationship between video content characteristics, the available bit rate, and their combined impact on perceived quality, this dissertation investigates quality-driven rate adaptation techniques that improve the final perceived quality in the context of two different video streaming application scenarios. First, the dissertation examines applications involving the transmission of multiple simultaneous multimedia streams to a single receiver, such as the transmission and display of multiple video streams depending on the application scenario. The problem of efficiently allocating the available session bandwidth among the individual multimedia streams is addressed, allowing participating streams to jointly adapt their transmission rates while considering their respective time-varying quality. Appropriate adaptation time scales are proposed, aligned with changes in video content, such as scene changes, along with an inter-stream adaptation mechanism that takes into account the time-varying objective quality of the participating streams. Experimental results demonstrate the benefits of the proposed method in terms of improved session quality and better utilization of available bandwidth, compared to: (a) priority-based inter-stream adaptation and (b) the case where session streams are transmitted over independent congestion-controlled connections. Next, the dissertation addresses the problem of providing smooth quality-driven rate adaptation for live, unicast real-time video streaming. Since real-time performance is required, an objective quality metric cannot be directly applied within the stream, as it is computationally demanding. For this reason, artificial neural networks are employed to predict quality scores in real time. The predictions are based on descriptive features of the video content and the bandwidth available to the stream. The limitations of existing approaches in providing stable or smooth quality are identified. A rate–quality controller based on fuzzy logic principles is then developed, aiming to reduce annoying short-term quality fluctuations caused by mismatches between available bandwidth and the rate required for stable quality. Based on the neural network’s quality predictions, the controller continuously monitors recent quality values, the nominal transmission rate, and the buffer occupancy levels of the participating buffers, in order to compute appropriate encoding rates that eliminate short-term quality variations. Experimental results show that the proposed solution provides significant short-term quality stability and smooths undesirable oscillations toward extreme low and high quality levels, while respecting rate constraints and maintaining buffer stability.By presenting extensive experimental results across a wide variety of video sequences, this dissertation demonstrates that video streaming systems can exploit objective perceptual quality metrics to deliver improved presentation quality by applying quality-driven adaptation techniques tailored to the semantics of each specific streaming application.

Degree

thesis:*
Grantor dc:publisher
Institutes outside Greece
Year dc:date
2004

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Miras, Dimitrios
  • Μοίρας, Δημήτριος

Subjects

dc:subject × 17

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/61869

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Miras, Dimitrios; Μοίρας, Δημήτριος. On quality aware adaptation of internet video. Institutes outside Greece, 2004. http://hdl.handle.net/10442/hedi/61869