Ti5216100 Machine Vision

1
Ti5216100Machine Vision

• MPEG
• Antti Hovi
• Teemu Kinnunen
• Jari Mononen
• Risto Westman
• 2006

2
Contents

• Introduction to MPEG
• Fundamentals of MPEG Video Compression
• Basic Concepts of MPEG (MPEG-1)
• MPEG-2
• MPEG-4
• Summary

2
3
Why is video compression needed?

• Example Titanic
• 3 h 15 min, 768x576 pixels, 8 bits / color, 25
  fps ? 360 GB
• HDTV resolution (1920x1080) ? 1695 GB
• Different needs
• DVD
• TV/HDTV broadcasting
• Mobile video

3
4
Why Is MPEG So Successful?

• Generic structure
• Wide range of applications
• Application-specific parameters
• Only the decoder structures and bit stream
  formats are standardized

4
5
MPEG Standards

• MPEG-1 MPEG-2 video and audio compression
  standards
• MPEG-2 also a transport standard
• MPEG-3 designed for HDTV
• MPEG-4 expands MPEG-1 and MPEG-2
• MPEG-7 for describing multimedia contents
• MPEG-21 is a multimedia framework

5
6
Fundamentals of MPEG Video Compression (1/3)

• Video sequences contain some redundancy
• Temporal redundancy (between frames, interframe)
• Spatial redundancy (inside frames, intraframe)
• Subjective redundancy

6
7
Fundamentals of MPEG Video Compression (2/3)
7
8
Fundamentals of MPEG Video Compression (3/3)

• Eyes are sensitive to brightness ? chrominance
  components can be subsampled
• Images divided into YUV channels
• Y Luminance
• U Chrominance (blue)
• V Chrominance (red)
• Subsampling with YUV ratio 411 or 422

8
9
Discrete Cosine Transform (DCT)

• Fast, O(Nlog2N)
• 8x8 pixels blocks

9
10
MPEG Frame Encoding
10
11
Differential Pulse Code Modulation (DPCM)

• Temporal redundancy using previous values
• 7 9 9 9 11 ? 7 2 0 0 2
• Used in macroblock content and in DC
  coefficients.

11
12
Motion Compensated Prediction
12
13
Group of Pictures (GOP)

• Video stream consist of sequences
• Sequences are divided in group of pictures

13
14
Macroblock

• Basic building block
• Consist of blocks
• Three different kind of blocks (luminance and
  chrominance)
• Each block is 8x8 pixels
• Conditional replenishment

14
15
I-, P-, B-pictures

• Intra-picture
• keyframe
• Predicted-picture
• Predicted from previous I- or P-picture
• Bidirectional-picture
• Predicted from previous and next I- or P-picture

15
16
I-, P-, B-picture (c.)
16
17
MPEG-2

• Advantages
• More efficient encoding when using interlaced
  video
• Higher resolutions (HDTV ready)
• More flexible encoding possibilities using levels
  and profiles
• Scalability

17
18
MPEG-2 Scalability

• SNR Scalability
• Signal-to-noise Ratio
• Spatial Scalability
• Multiple layers
• Temporal Scalability
• Adds frames in the video stream
• Data Partitioning
• Like JPEGs progressive mode

18
19
MPEG-2 Profiles and Levels

• Profile sets quality of the video
• Five profiles Simple profile, Main profile, SNR
  Scalable Profile, Spatially scalable profile and
  High profile
• Level sets resolution of the video
• Four levels Low, Main, High-1440 and High
• Different profiles has different levels
• Main profile has four levels
• Simple level has only one level

19
20
Differences between MPEG-1 and MPEG 2

• MPEG-2 Is superset of MPEG-1 with couple
  different data elements
• IDCT mismatch handling
• MPEG-1 and MPEG-2 places differently chrominance
  samples
• MPEG-1 has fixed picture rate
• In MPEG-1 pictures indexes differs
• Pictures information is placed differently

20
21
MPEG-4 Standard definitions

• Systems (Transport stream)
• Visual (Such as codecs)
• Audio
• Conformance
• Reference software
• Delivery Multimedia Integration framework
• Optimized reference software
• Carriage on IP networks
• Reference hardware
• Advanced Video coding
• Scene description and application engine

• ISO Base Media File format
• Intellectual property management and protection
• MPEG-4 File format
• AVC File format
• Animation Framework Extension
• Timed Text subtitle format
• Font compression and streaming
• Synthesized Texture stream
• Lightweight Scene representation
• MPEG-J Graphical Framework Extension
• Open Font Format Specification

21
22
MPEG-4 Standard

• What MPEG-4 adds to previous standards
• Object framework
• Interactive content
• Why was MPEG-4 standard developed?
• coded representation of natural and synthetic
  data
• Scalability to several transport methods
• Main improvement techniques in video quality
• Image blocks are arbitrary shaped
• Sprite coding technology
• YV12 Color space (emphasis on Luma field)

22
23
Summary

• Redundancies and Transform-Domain Coding
  essential in video compression
• DCT, subsampling
• Motion prediction
• MPEG is generic, enables a wide range of
  applications
• Video compression principles remain basically the
  same throughout the standard family

23
24
References

• Sikora T. 1997. MPEG Digital Video-Coding
  Standards. IEEE Signal Processing
• Magazine, vol. 14, issue 5, pages 82-100. ISSN
  1053-5888
• Jain A. K. 1989. Fundamentals of Digital Image
  Processing.
• Englewood Cliffs Prentice-Hall. 569 p. ISBN
  0-13-336165-9
• Mitchell J, Pennebaker W, Fogg C, LeGall D. 1996.
  MPEG Video Compression Standard.
• Chapman Hall. 470 p. ISBN 0-412-08771-5
• A Beginners Guide for MPEG-2 Standard. Available
  
• http//www.fh-friedberg.de/fachbereiche/e2/telekom
  -labor/zinke/mk/mpeg2beg/beginnzi.htm.
• Referenced 2006-10-18
• MPEG-2 Overview. Available http//erg.abdn.ac.uk/
  research/future-net/digital-video/mpeg2.html.
• Referenced 2006-10-18

24