Image and Video Compression
                  EE398A

                 Bernd Girod
       Information Systems Laboratory
 ...
Image and Video Compression Everywhere …
   Fax machines
   Digital still cameras
   Digital camcorders
   Digital tel...
Motivating Image Compression
   Binary image (facsimile Group 3)
       8.5 x 11 in document scanned at 7.7 lines/mm (“f...
Motivating Video Compression
    Digital video studio standard ITU-R Rec. 601

                                       Y  ...
How does compression work?
   Exploit statistical redundancy
       Take advantage of patterns in the signal
       Des...
Visual Redundancy
   For images to be viewed by humans, no need to represent
    more than the visible resolution in
    ...
Typical Image Coder and Decoder
                                                                                          ...
Discrete Wavelet Transform




Bernd Girod: EE398A Image and Video Compression   Introduction no. 8
Example: JPEG 2000 Compression
uncompressed: 786Kbytes                   75:1 compression (10.6 Kbytes)




   512 x 512 P...
EE398A Outline
1.   Basic building blocks
     1.   Entropy coding: Huffman codes, Golomb codes, arithmetic coding
     2....
EE398A People




   Instructor                  Teaching Assistant
                                                    Co...
EE398A Organisation
   Class home page: http://www.stanford.edu/class/ee398a
   Homeworks
       7 problem sets, requir...
EE398A Term Projects - General
   Work in groups of 2-3 students, 50 hours per person
   Submit project proposal by Febr...
ISE laboratory
   Created by equipment grants from Hewlett-Packard, Xerox,
    and Intel
   Exclusively a teaching labor...
Review & Further Reading
   Selected papers posted on Web site
   Slides available as hand-outs and as pdf files on
    ...
Upcoming SlideShare
Loading in...5
×

Image and Video Compression

3,869

Published on

0 Comments
2 Likes
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total Views
3,869
On Slideshare
0
From Embeds
0
Number of Embeds
1
Actions
Shares
0
Downloads
95
Comments
0
Likes
2
Embeds 0
No embeds

No notes for slide

Image and Video Compression

  1. 1. Image and Video Compression EE398A Bernd Girod Information Systems Laboratory Department of Electrical Engineering Stanford University Winter 2009/10
  2. 2. Image and Video Compression Everywhere …  Fax machines  Digital still cameras  Digital camcorders  Digital television broadcasting  Digital cinema  Digital video/versatile disk (DVD)  Personal video recorder (PVR, aka TiVo)  World Wide Web  Internet video streaming  Video conferencing  ... Bernd Girod: EE398A Image and Video Compression Introduction no. 2
  3. 3. Motivating Image Compression  Binary image (facsimile Group 3)  8.5 x 11 in document scanned at 7.7 lines/mm (“fine mode”), 1664 pixels/line, with 1 bit/pixel  3.255 Mbits for 1 page = 5.65 minutes over 9600 baud connection  Color photo  35 mm film scanned at 12 resolution (3656x2664 pixels)  Corresponds to digital still camera with 10Mpixel sensor  With 3 color components (RGB) and 8bits/color: 233 Mbits total  For digital cinema: 24 times per seconds Bernd Girod: EE398A Image and Video Compression Introduction no. 3
  4. 4. Motivating Video Compression  Digital video studio standard ITU-R Rec. 601 Y Cb Cr Sampling rate 13.5 MHz 6.75 MHz 6.75 MHz Quantization 8 bit 8 bit 8 bit Raw bit rate 216 Mbps w/o blanking intervals 166 Mbps  Some interesting bit-rates (some evolving with time)  Terrestial TV broadcasting channel ~20 Mbps  DVD (4.7…17 GB/length of movie) 5...20 Mbps  Ethernet/Fast Ethernet <10/100/1000 Mbps  DSL downlink 384...2048 kbps  Wireless cellular data 9.6...384 kbps Bernd Girod: EE398A Image and Video Compression Introduction no. 4
  5. 5. How does compression work?  Exploit statistical redundancy  Take advantage of patterns in the signal  Describe frequently occurring events efficiently  Lossless coding: only statistical redundancy  Introduce acceptable deviations  Omit information that the humans cannot perceive  Match the signal resolution (in space, time, amplitude) to the application  Lossy coding: exploit both visual and statistical redundancy Bernd Girod: EE398A Image and Video Compression Introduction no. 5
  6. 6. Visual Redundancy  For images to be viewed by humans, no need to represent more than the visible resolution in  space  time  brightness  color  Required resolution might depend on image content (“masking”)  For some applications, only a specific region of the image might be relevant, e.g., in  medical imaging  surveillance Bernd Girod: EE398A Image and Video Compression Introduction no. 6
  7. 7. Typical Image Coder and Decoder Encoder image x Transform coefficients y Quantizer indices q Entropy coder bit-stream c yT x  qQ y  cC q reconstructed ˆ image x inverse ˆ coefficients y dequantizer Entropy indices q transform decoder x  T 1  y  ˆ ˆ y  Q 1  q  ˆ  q  C 1 c Decoder  Transform T(x) usually invertible  Quantization Q  y  not invertible, introduces distortion Combination of encoder C  q  and decoder C  c  1  lossless Bernd Girod: EE398A Image and Video Compression Introduction no. 7
  8. 8. Discrete Wavelet Transform Bernd Girod: EE398A Image and Video Compression Introduction no. 8
  9. 9. Example: JPEG 2000 Compression uncompressed: 786Kbytes 75:1 compression (10.6 Kbytes) 512 x 512 Pixels, 24-Bit RGB Bernd Girod: EE398A Image and Video Compression Introduction no. 9
  10. 10. EE398A Outline 1. Basic building blocks 1. Entropy coding: Huffman codes, Golomb codes, arithmetic coding 2. Quantization, Lloyd algorithm, high-rate performance 3. Predictive coding 2. From transform coding to JPEG 1. Orthonormal transforms, coding gain, KLT, separability, DCT 2. Baseline JPEG, quantization matrix, run-level coding 3. From wavelets to JPEG-2000 1. DWT, perfect reconstruction filter bank, 5/3 wavelet 2. JPEG-2000, EBCOT, embedded quantization, bit-plane coding 4. From motion compensation to MPEG and H.264 1. M.c. prediction, hybrid coding, block matching, subpel accuracy 2. Video coding syntax, macroblocks, GOPs, MPEG1/2, H.264/AVC Bernd Girod: EE398A Image and Video Compression Introduction no. 10
  11. 11. EE398A People Instructor Teaching Assistant Course assistant Bernd Girod Mina Makar Office hours: Office hours: Kelly Yilmaz Fr, 1:30-3 pm Mo, 4-6pm Packard 259 Packard 373 Packard 021 Bernd Girod: EE398A Image and Video Compression Introduction no. 11
  12. 12. EE398A Organisation  Class home page: http://www.stanford.edu/class/ee398a  Homeworks  7 problem sets, require computer + Matlab (Image Proc. Toolbox)  Handed out Tuesdays, due one week later, 2:15 p.m.  Late homework policy  0-24 hours – subtract 10%  24-48 hours – subtract 25%  48-72 hours – subtract 50%  After 72 hours – subtract 100%  Grading  Homeworks: 50%  Term project: 50%  No mid-term, no final Bernd Girod: EE398A Image and Video Compression Introduction no. 12
  13. 13. EE398A Term Projects - General  Work in groups of 2-3 students, 50 hours per person  Submit project proposal by February 4  Class-room presentations of projects: March 9/11  Written report due: March 11  Presentations and reports will be posted online  Project grade based on  Technical quality, significance, and originality of results 50%  Written report 25%  Class-room presentation 25% Bernd Girod: EE398A Image and Video Compression Introduction no. 13
  14. 14. ISE laboratory  Created by equipment grants from Hewlett-Packard, Xerox, and Intel  Exclusively a teaching laboratory  Location: Packard room 021  20 Linux PCs, 2 Windows PCs, scanners, printers etc.  Access:  Door combination for lab entry will be provided by TA  Account on ise machine will be provided to all enrolled in class Bernd Girod: EE398A Image and Video Compression Introduction no. 14
  15. 15. Review & Further Reading  Selected papers posted on Web site  Slides available as hand-outs and as pdf files on the web  Lecture videos online as part of alpha test of IRoI streaming system (no guarantees)  Books (recommended, but not required)  D. S. Taubman, M. W. Marcellin, JPEG2000 – Image Compression Fundamentals, Standards, and Practice, Kluwer Academic Publishers, 2002.  Y. Wang, J. Ostermann, Y.-Q. Zhang, Video Processing and Communications, Prentice-Hall, 2002. Bernd Girod: EE398A Image and Video Compression Introduction no. 15
  1. A particular slide catching your eye?

    Clipping is a handy way to collect important slides you want to go back to later.

×