CHAPTER 7 MPEG1 and MPEG2

1
CHAPTER 7MPEG-1 and MPEG-2
2
OUTLINE

• Introduction
• The MPEG-1 Video Standard
• MPEG-1 Implementation Issues
• The MPEG-2 Video Standard
• The MPEG-4 Coding Standard
• Future Directions MPEG-7

3
Introduction
4
Video/Image Standard Applications
5
Digital Video Formats
6
Progressive Interlace
7
MPEG

• MPEG Moving Picture Experts Group
• MPEG includes
• MPEG-1,MPEG-2,MPEG-4,MPEG-7, MPEG-21
• MPEG-1 Standard
• 11172-1 Systems
• 11172-2 Video
• 11172-3 Audio
• 11172-4 Conformance
• 11172-5 Software

8
MPEG-1 Standard

• Media storage
• Bitrate Up to about 1.5Mbps
• Completed in 1992
• VCD,MP3
• MPEG-1 does not recognize interlaced sources

9
Compression Functions within MPEG

• Sample rate reduction in the spatial and temporal
  domains of both the luminance and chromance
  components,
• Block-based DCT for the intraframes and
  interframes,
• Block-based motion compensation for predictive
  and interpolative interframes, and
• Huffman coding for the lossless compression of
  motion vectors and the quantized DCT
  coefficients.

10
Definition of a Macroblock in MPEG-1
11
Picture Types

• Intra-pictures (I-pictures) are compressed using
  intraframe coding that is, they do not reference
  any other pictures in the coded bitstream.
• Predicted pictures (P-pictures) are coded using
  motion-compensated prediction from past
  I-pictures or P-pictures.
• Bidirectional predicted pictures (B-pictures)
  provide the highest degree of compression. They
  are coded using motion-compensated prediction
  from either past and/or future I-pictures or
  P-pictures.

12
Picture Types (Cont.)

• DC-coded pictures (D-pictures) are similar to
  I-pictures however, only the DC coefficients
  from the DCT outputs are present.

13
Bit Distribution among I-, P-, and B- Pictures
160
140
120
Picture size (kbits)
100
80
60
40
20
0
0.00 50.00 100.00
150.00 Picture number (in encoding order)
14
MPEG Video Encoder

• Preprocessing
• Color conversion
• Format translation
• Prefiltering
• subsampling

15
Coding of I-Picture
16
Coding of I-Pictures

• Intra-frame coding
• 8x8 2D DCT
• Differential coding of DC coefficients
• Uniform quantization
• Run length coding of zeros with zig-zagscan
• Entropy coding

17
Coding of P-Pictures
18
Coding of P-Pictures

• Motion compensated prediction from I-or
  P-pictures
• One motion vector per microblock
• Coding of prediction error with 8x8 DCT,
• uniform quantization, zig-zag scan (like
  P-pictures)

19
Coding of B-Picture
20
Coding of B-Pictures

• For B-pictures, the motion estimation process is
  performed twice, once for a past picture and once
  for a future picture, so as to yield two motion
  vectors.
• The encoder can form a prediction error
  macroblock from either of the two candidate
  macroblocks or from their average.
• In MPEG terminology, this is referred to as
  interframe interpolative coding.
• The prediction error is then coded using the
  block-based DCT. The quantized DCT coefficients
  of the prediction errors, together with the
  motion vectors are then multiplexed and coded
  using the variable-length coder.
• Unlike JPEG, MPEG uses only a Huffman coder for
  variable-length coding.

21
Coding of B-Pictures (Cont.)

• Because of the bidirectional prediction, some
  reordering of the input pictures is needed at the
  encoder so that they are delivered in the correct
  order to the decoder.
• In the previous example, the correct encoding
  sequence is p1, p4, p2, p3, p7, p5, p6, and p8.
• The decoder will need to reorder these pictures
  so that they can be displayed in the correct
  order.

22
Bidirectional Motion Compensation
Past picture
Best matching macroblock
Current picture
MV-1 (x,y)
MV motion vector
Future picture
(x, y)
(x, y) MV-2
-
-



Forward prediction error


Best matching macroblock
Interpolative prediction error

Backward prediction error
1/2
Prediction error
To DCT
MUX
23
Structure of the Coded Video Bit Stream
24
MPEG Video Decoder
25
MPEG-2 Standard

• The range of coding support provided by MPEG is
  divided into profile and levels.
• A profile is a defined subset of the entire
  bitstream syntax specified by MPEG-2. The
  profiles are Simple, Main, 422, SNR, Spatial,
  High, and Multiview.
• Within a profile, a level is defined as a set of
  constraints imposed on the parameters of the bit
  stream, such as picture resolution or maximum bit
  rate.
• For each profile, the four levels are Low (for
  SIF resolution pictures, Main (for CCIR 601
  resolution pictures, High-1440, and High (for
  HDTV resolution pictures)

26
MPEG-2 Standard (Cont.)

• MPEG-2 is backward compatible with MPEG-1 that
  is, an MPEG-2 decoder should be able to decode an
  MPEG-1 coded stream.
• MPEG-2 suppotrs both interlaced and noninterlaced
  pictures.
• In addition to the 420 format used in MPEG-1,
  MPEG-2 also supports the 422 and 444 color
  subsampling formats.
• When the coded bit stream has no scalable
  extensions, then all blocks within a macroblock
  are 8 ? 8. With scalable extensions, macroblocks
  may contain scaled blocks with lower resolutions,
  such as 1 ? 1, 2 ? 2, or 4 ? 4.

27
MPEG-2 Standard (Cont.)

• The structure of the luminance macroblocks is
  different for frame DCT coding and field DCT
  coding.

28
Prediction Modes and Motion Compensation

• In MPEG-2, the picture sequence can be either a
  collection of frame pictures or a collection of
  field pictures.
• Two classes of prediction are supported, namely,
  frame prediction and field prediction.

Top Field
TOP Field
TOP Field
First Field
Bottom field of frame
Bottom Fields
Bottom Fields
Second Field
Reference frame
Field Prediction
29
Field and Frame Prediction

• In MPEG-2, unlike, all motion vectors are
  specified at half-pixel resolution.
• Let (u, v) denote the motion vector of the
  luminance components in a macroblock Due to the
  subsampling of the chroma components, their
  motion vectors are given by(u/2, v/2) for the
  420 format and by (u/2, v) for the 422
  format.
• For the 444 sampling format, the chroma
  components use the motion vectors of the
  luminance without any scaling.

30
Field and Frame Prediction (Cont.)

• To improve the prediction efficiency, MPEG-2
  provides two additional motion compensation
  modes.
• 16 ? 8 motion-compensation mode.
• Dual-prime motion compensation