ECE 434 – Multimedia Networks
Spring 2009
David Swiston
MPEG2 Computer Project
1. Introduction
The following report and accompanying code detail an investigation into the MPEG-2 compression
algorithm. The steps and theory involved in encoding and decoding the MPEG-2 bit stream will be
discussed. A MATLAB based simulation of the encoding and decoding process is provided along with
experimental results using the MATLAB simulation.
2. MPEG-2 Algorithm
1. Caveat
This report essentially ignores the construction and framing of the MPEG-2 bitstream. This project
need not concern itself with this due to the assumption that the encoding and decoding is done over a
direct communication line. There is no need to packetize and frame the information. Ignoring this
aspect of MPEG-2 still allows the important concepts of MPEG-2 coding to be analyzed and discussed.
2. Introduction and History
The MPEG-2 algorithm is a lossy compression algorithm intended to remove spacial and temporal
redundancy in a video in order to ease transmission and storage requirements for videos. Using a
simple example, a 1280 horizontal by 780 vertical pixel, 24 bit-per-pixel video at 24 frames per second
would require 530,841,600 bits-per-second of bandwidth to transmit with a two hour movie requiring
477,757,440,000 bytes of storage.
1280 pixels∗720 pixels∗24 bits /pixel∗24 frames/s=530,841,600 bits /s
530,841,600 bits /s∗60 s/min∗60 min /hr
8bits /byte =477,757,440,000 bytes
Given the very large requirements to transmit a digital video as described above, it is necessary to
compress the video to make storage and transmission practical.
MPEG-2 is widely used in many applications, allowing video transmissions such as the example
provided above feasible. Its development began in 1990, was approved in November of 1994, and the
final standard published in 1995. The initial application was as a standard for digital broadcast TV,
offering higher target bit rates than the MPEG-1 standard and also supporting interlaced video. It has
since been widely used in many applications such as in DVDs, ATSC, and DVB.
3. Colorspace
The MPEG-2 standard uses the YCbCr colorspace in order to take advantage of the Human Visual
System (HVS). The Y component in YCbCr represents the luminance information and Cb and Cr
represent chrominance differences. The eye is less sensitive to chrominance than it is to luminance.
MPEG-2 exploits this fact by allowing for chrominance subsampling. Subsampling schemes 4:4:4,
4:2:2, and 4:2:0 are used. The format 4:4:4 keeps all of the chrominance information, 4:2:2 keeps half
of the chrominance information by horizontally subsampling by a factor of two, and 4:2:0 keeps a
quarter of the chrominance information by subsampling both the horizontal and vertical dimensions by
two. By removing chrominance information, either of the later two formats will automatically result in
a data reduction with little to no perceptible difference to the eye. The 4:2:0 format is the most
pervasive subsampling format. The MATLAB simulation and subsequent discussions assume the 4:2:0
format.
