MPEG | Sahithyan's Notes

MPEG stands for Moving Picture Experts Group, which develops standards for coded media and associated systems.

MPEG Standards

Standard	Scope and Examples
MPEG-1	VCD audio and video, with a target around 1.5 Mbps at 352 × 240 for the lecture’s example.
MPEG-2	DVD and digital television, including interlaced video and different profiles and levels.
MPEG-3	HDTV proposal discontinued when MPEG-2 was considered sufficient.
MPEG-4	Audio-visual objects, 3D content, low-bit-rate coding, and rights-management support.
MPEG-4 Part 10	H.264, also called Advanced Video Coding (AVC).

The lecture’s technical-specification breakdown refers to MPEG-2.

MPEG-2 Part	Purpose
1	Systems, synchronisation, and multiplexing.
2	Video coding.
3	Audio coding.
4	Conformance testing.
5	Software simulation.
6	Digital Storage Media Command and Control (DSM-CC).
7	Advanced Audio Coding (AAC).
8	Discontinued part.
9	Real-time interfaces.
10	DSM-CC conformance extensions.

Classic MPEG spatial processing follows a JPEG-like approach.

Classic MPEG motion prediction uses macroblocks with a 16 × 16 luminance region and associated chrominance data.

Frame	Coding Method	Typical Relative Size in the Lecture
I-frame	Intra-coded without another frame as a reference.	Largest.
P-frame	Predicted from a preceding I- or P-frame.	Approximately 1/3 of an I-frame.
B-frame	Can use preceding and following I- or P-frames.	Approximately 1/6 of an I-frame.

Relative sizes depend on content and encoder settings.

A group of pictures (GOP) is a sequence of coded pictures organised around reference pictures.

A closed GOP requires no reference pictures outside the group.

Mode	Allocation	Implication
CBR	Constrains output to a target rate.	Simplifies delivery over a fixed-capacity channel.
VBR	Allocates more bits to complex scenes.	Improves allocation across content with changing complexity.

Stored media can use 2-pass VBR encoding to analyse scene complexity before allocating bits.

Standard	Properties
MPEG-1 audio	Defines Layers I, II, and III with different coding complexity.
MPEG-1 Layer III	MP3, with sampling rates of 32, 44.1, or 48 kHz and bit rates from 32 to 320 kbps.
MPEG-2 audio	Extends support to multichannel audio and lower sampling frequencies.
AAC	Supports sampling frequencies from 8 to 96 kHz and improved coding efficiency.

MPEG audio can use CBR or VBR.

Container	Intended Environment
Program Stream (PS)	Relatively reliable media such as disks.
Transport Stream (TS)	Error-prone delivery such as broadcast or network links.

Presentation timestamps (PTS) specify when decoded media should be presented.

PTS uses a 90 kHz clock scale.
Program Clock Reference (PCR) values let a transport-stream receiver recover sender timing.
A phase-locked loop (PLL) can synchronise the receiver clock.

Adapted from December 2022 and November 2023, Question 4. The sequence below is an illustrative closed GOP.

Display order:

I0  B1  B2  P3  B4  B5  P6

Assume each P-frame references the preceding I- or P-frame. Each B-frame uses both surrounding reference frames.

A valid decoding order is:

I0  P3  B1  B2  P6  B4  B5

To seek directly to B4 with an empty decoder:

B1 and B2 are unnecessary for this seek because no later frame references them in this example.

Losing P3 affects P6 and every B-frame that depends on either reference.
Losing B4 does not invalidate the other frames in this example.
Longer GOPs can reduce I-frame overhead but increase the reference chain needed for seeking and recovery.
Only a closed GOP avoids references to pictures outside the group.