Skip to main navigation Skip to search Skip to main content

Codage vidéo basse complexité sans perte de qualité visuelle supportant les résolutions HD sur le DSP asynchrone Opus d'Octasic

Translated title of the thesis: Low complexity video coding for visually lossless compression supporting HD résolutions on Octasic's Opus asynchronous DSP
  • Louis Poudrier--Racette

Student thesis: Master's thesisMaster in Engineering: Engineering

Abstract

This research thesis presents a new method for video coding, which optimizes the transfer of video data within the Octasic Vocallo MGW system. This method is presented as a solution to the problems posed by the exchange of large amounts of data generated by a system in which a high-definition video content is present in an uncompressed form. More specifically, this work addresses the problem of transferring video data over a Gigabit Ethernet link between the cores of an Octasic silicon chip. The Vocallo MGW application runs on Octasic’s asynchronous DSP. This technology imposes several constraints related to the selection of instructions and memory management. The compression method developed in this research project aims to satisfy three constraints. The first one is the compression ratio required to enable bidirectional communication over a Gigabit Ethernet link. The second one is to preserve the visual quality. The third one is the execution time since the solution must operate in real-time on Octasic’s Vocallo MGW processing system. The proposed compression method is based on the HACP-SBT algorithm available from the literature but contains several key improvements to make it functional in the context of the Vocallo MGW. This work’s contributions include the simplification of the prediction algorithm, the adaptation of the coding syntax to allow packetisation and the definition of a flexible packet header, and the creation of three coding modes to satisfy the constraints. These three modes include a lossless coding mode supplemented by two lossy compression modes : the calculated lossy mode promoting visual quality and the rapid lossy mode favoring lower execution times. The analysis of these three coding modes shows that all the modes meet the constraint of compression ratio and visual quality. However, only the processing of small resolutions, such as QCIF, is fast enough to meet the real-time constraint.
Date18 Feb 2014
Original languageFrench
Awarding Institution
  • École de technologie supérieure
SupervisorStéphane Coulombe (Supervisor)

Cite this

'