Combining Alm Codecs For Improved Video Streaming

Combining ALM and customized codecs in

improving video streaming

AbstractвЂ"Streaming media and IPTV are two

emerging technologies that promise to bring high

quality video signals to the end user. IPTV is likely

to use application layer multicast to distribute the

video to different users interested in the same content.

Though much work has been done on application layer

multicast and how the receivers organize themselves in

a tree structure, not much prior work has been done on

the interaction between video codecs and application

layer multicast trees.

In this paper we are going to study the interaction

of videocodecs with application layer multicast trees.

First, we study the effect of the codec parameters on

the quality of the video signal that the end user experiences.

Second, based on the result of the comparision,

we study how reorganization of the tree structure can

improve the end user experience. To evaluate end user

experience we use the PSNR metric.

I. INTRODUCTION

Video is a high bandwidth signal which is susceptible

to delay and packet loss. Different methods for

distributing video while accomendating these challenges

have been proposed. These methods can be divided

into three classes; Native multicast, application

layer multicast and peer to peer based distribution.

The former relies on the physical routers of the

internet to support native multicasting, which, as of

today, has unfortunetly not been widely deployed.

Peer to peer based distribution has been deployed

in some commercial video streaming applications

using technologies like Bittorrent to distribute the

video1. In the scientific community the main focus

for distributing video have been on application layer

multicast.

Though much work has been done on application

layer multicast and how the receivers organize

themselves in different topologies, not much prior

1http://en.wikipedia.org/wiki/P2PTV

work has been done on the interaction between video

codecs, user experience and how these topologies are

created.

The focus of this paper is on how to construct

data distribution trees with properties that are well

suited for video distribution.We have chosen to focus

on the MPEG-4 video codec for encoding the test

videos. The reason for choosing MPEG-4 is that it

is widely deployed and used in the real world, which

gives us different well documented methods for comparing

video qualities. More specificly we are going

to use the PSNR2 as comparison metric between both

video streams and between tree topologies.

Data distribution trees can experience error propagation

if proper messures are not taken. That is, as

the video content is distributed down the layers of

the tree, there is a potential for propagation of errors

caused by missing packets. This may lead to some

nodes in the tree not receiving a proper video signal.

In section ?? we manipulate different parameters

of the MPEG-4 video codec to see which are best

suited for internet-like video distribution. We will

also look at decoding and encoding of the received

video content in different resolutions and with different

bit-rates before forwarding the video signal

to the lower layer nodes of the tree that have less

bandwidth capacity. In section ?? we use this result

to test the quality of different distribution trees, using

the PSNR at each end host to compare the trees. We

are constructing NICE trees having different sized

clusters and different metrics for joining hosts.

II. BACKGROUND

A. Application layer multicasting

As mentioned in section I we are going to focus on

application layer multicasting as means for distributing

the video stream. Application layer multicasting

2http://en.wikipedia.org/wiki/PSNR

2

differes from native or network layer multicasting in

where the responsibility for duplicating the stream

lies. In application layer multicasting it is the end

host, which has the responsibility, which results in

many