Quality aware transcoding: An application level technique to dynamically adapt multimedia content

The web is emerging as the primary data dissemination and e-commerce mechanism. Users access the Internet from a wide variety of devices with different resource constraints in terms of their processing, storage, network and display capabilities. Users are not only accessing the Internet from traditional desktops, but they are also using mobile devices such as palmtops and laptops as well as newer consumer devices such as webtops and navigation systems. Popular web servers also experience periods of bursty load from flash crowds. The problem of slow and expensive networks to the last hop as well as bursty web traffic is compounded by the proliferation of inexpensive devices that can create, share and consume multimedia objects. A wider audience is encouraging e-commerce sites to use multimedia objects extensively. Multimedia objects, which tend to be large, are placing enormous strains on the network infrastructure. Web services need to dynamically manage available computation and network resources because of these highly bursty and demanding access patterns. The problem we are attempting to solve is caused by the combination of the emphasis upon multimedia objects on the web, the wide range of resources of the access devices and the variability in network bandwidth available.;In this dissertation, I present image transcoding as an application level technique to dynamically customize content. Transcoding is a promising technology that allows systems to effect a quality-versus-size tradeoff on multimedia objects. Transcoding can be used to dynamically customize objects for the prevailing operating environment. First, I analyze the characteristics of web images to choose appropriate transcoding technologies. Next, I quantify the quality versus size trade-off characteristics of a promising transcoding algorithm. I then use this informed technique in a web service for providing bandwidth consumption control and for providing differentiated Quality of Service. I also use this technology in a web service that customizes objects for clients accessing the web using various network technologies. This technology also allows a multimedia capture device (such as a digital camera) to manage its constrained storage and battery power resources. Through implementations of informed transcoding in a popular web server, I demonstrate how I can provide "appropriate service" for the current constrained operating environment. I show that by understanding the image characteristics, informed transcoding avoids unnecessary operations that lose quality as well as place unnecessary loads on the infrastructure.