IBC2024 Tech Papers: Novel coding methods for storage bit-cost, transcoding complexity, and transmission efficiency trade-off optimization of multi-profile video delivery system

Abstract

This paper addresses the problem of multi-profile encoding and delivery system optimization for the purpose of standard HTTP-based Adaptive Bitrate (ABR) video streaming. Such delivery systems must process, encode, and usually store the same content in different (bitrate, resolution) pairs, which defines a set of encoding profiles or coded representations (a.k.a. bitrate ladder), to serve and adapt the video content to various end user bandwidth requirements and device capabilities. The presented research work specifically targets such a system to improve the trade-off between the storage bit-cost of the different representations, the transcoding complexity and transmission efficiency (i.e. bitrate-quality trade-off at transmission) of the requested representation by the end-client while guaranteeing that the delivered output bitstream remains compliant with the legacy decoding system available at the client. For that purpose, a joint multi-profile coding format with corresponding fast transcoding method is proposed and assessed against State-of-the Art methods.

Introduction

Video streaming services heavily rely on HTTP-based Adaptive Bitrate (ABR) streaming technologies, such as Dynamic Adaptive Streaming over HTTP [1] or HTTP Live Streaming (HLS) [2], to serve video content to varying end-user device capabilities and network conditions. To adapt to the end-client request, ABR delivery system must commonly process, encode, and store the same video content in different resolution and bit-rate pairs, which define a set of encoding profiles or coded representations (a.k.a. bitrate ladder). As a first path of system optimization the bitrate ladder can be optimized per content, using traditional [3-5] or machine learning [6-8] approaches, and with knowledge of the video coding standard or codec efficiency in use [9]. Complementary, and by considering the signal redundancy between the representations, we investigate and introduce novel joint coding formats and transcoding methods to optimize the trade-off between storage cost, transcoding complexity, and transmission cost of those representations in a multi-profile video coding system for standard ABR streaming. The two most common approaches for multi-profile video delivery are Simulcast (SC) and Full Transcoding (FT) which set two extremes in terms of optimization criteria. Simulcast offers the lowest transcoding complexity (i.e. highest scalability to profile requests) and best transmission efficiency but requires the largest amount of storage, while FT has the lowest storage cost but the highest transcoding cost and transmission bitrate overhead.