JPEG 2000

Image Compression Standard

GTDIR

Grupo de Investigación: Tratamiento Digital de Imágenes Radiológicas Enrique Nava, University of Málaga (Spain) Brasov, July 2006

Standard organization

Where does MPEG fit into the scheme of things

ISO Internation Standards Organization IEC International Electrotechnical Commission ISO/IEC JTC 1 Joint Technical Committee 1

Standard organization

Where does MPEG fit in the scheme of things ISO/IEC

Geneva, CH ITTF International Technical Task Force Geneva, CH

JTC 1

Joint Technical Committee - Information Technology Secretariat - ANSI New YorK City

SC xx

Subcommittee

SC 29

Subcommittee 29 Secretariat - ITSJC/IPSJ Tokyo, Japan

WG 1

JPEG

WG11

MPEG Moving Picture Experts Group Milan, IT

WG12

MHEG (last meeting 2001/03

How a standard is made?

   

Exploration

 (6-12 months) The search for new technology   Seek Industry experts Open seminars

Requirements

(6-12 months)  Establish the scope of work  Call for Proposals

Competitive phase

(3-6 months)  Do Homework   Response to CfP Initial technology selection

Collaborative phase

 Core Experiments (1 year)  Working Drafts

Standard phases

Document type 3 months Specification CD Amendment PDAM FCD FPDAM 4 months 2 months FDIS FDAM Corrigenda X Technical Report PDTR DCOR X X DTR

All documents start out in life as working drafts (WD)

publish IS AMD COR TR

Nomenclature

 WD - Working Draft      CD - Committee Draft FCD - Final Committee Draft FDIS© - Final Draft International Standard IS © - International Standard © indicates ISO copyright     PDAM - Proposed Draft Amendment FPDAM - Final Proposed Draft Amendment FDAM © - Final Draft Amendment AMD © - Amendment   DCOR - Draft Corrigenda COR © - Corrigenda    PDTR - Proposed Draft Technical Report DTR © - Draft Technical Report TR © - Technical Report  DoC - Disposition of Comments

Why JPEG 2000?

To improve quality-performance ratio:    Low bit-rate compression (<0.25 bpp) Lossy and lossless compression (both high performace) Computer graphics, bi-level and raster images    Transmission in noise channels Random access to data and progressive transmission Open architecture (different application can be optimized)

Low-rate transmission rate

0.25bpp (96:1) 0.125bpp (192:1) 0.5bpp (48:1)

Progressive transmission

Applications



Internet



Cellular phone imaging



Printers and scanners



Digital photography



Medical imaging



Digital libraries



E-commerce



Facsimile transmission

JPEG 2000: a big standard

 Core standard was proposed to finish in Dec 2000.

 It replaces JPEG standard (ISO 10918)  It has been proposed 12 parts  Part 1 is the core standard  Part 2 is extensions to the core system  Part 3 is motion Jpeg 2000  Parts 4-5 are conformance and reference software  Parts 6 and 12 are File Format issues  Parts 8-11 are applied extensions

JPEG 2000 parts (ISO/IEC 15444)

10 11 12 13 4 5 6 7 8 9 Part 1 2 3 Name Core Coding system Extensions Motion JPEG 2000 Conformance Reference software Compound image file format JPSEC: Secure JPEG 2000 JPIP: Interactivity tools, APIs and protocols JP3D: Extensions for 3D data JPWL: Wireless ISO Base Media File Format Entry level JPEG 2000 encoder IS Year 2000 2002 2002 2003 2003 2005 2004 2006 2005 2004

Building blocks

Color coding

 Irreversible Color Transform (ICT)  Used for conventional YCbCr transform for image and video signals.

 Not valid for lossless coding    

Y C b C r

          0 .

299 0 .

169 0 .

500 0 .

587  0 .

331  0 .

419    

R G B

       1 .

0   1 1 .

.

0 0 0 .

0000  0 .

3441 1 .

7718  0 0 0 .

.

.

114 500 081        

R G B

     1 .

4021 0 .

7142 0 .

0        

Y C C b r

   

Color coding



Reversible color coding

 Only integer operations Forward RCT: Y= (R+2G+B)/4 C b =B-G C r =R-G Inverse RCT: G= Y-(C b +C r )/4 B=C b +G R=C r +G

Discrete Wavelet Transform

      Multirresolution image representation: DWT.

The full-frame nature of the transform decorrelates the image across a larger scale and eliminates blocking artifacts at high compression.

Use of integer DWT filters allows for both lossless and lossy compression within a single compressed bit-stream.

DWT provides a frequency band decomposition of the image where each subband can be quantized according to its visual importance.

Two filters in Part 1: irreversible Daubechies (9,7) and reversible (5,3) Part 2 allows arbitrary filters

2-D Wavelet Decomposition

•Most wavelet based image compression systems use a class of analysis/synthesis filters known as

bi-orthogonal

filters: •– The basis functions of

h0

(

n

) and

g1

(

n

) are orthogonal as the basis functions for

h1

(

n

) and

g0

(

n

) are orthogonal.

•– Linear-phase (symmetrical) and perfect reconstruction.

•– Unequal length; odd-length filters differ by an odd multiple of two (e.g., 7/9), while even-length filters differ by an even multiple of two (e.g., 6/10).

•– Symmetric boundary extension.

Quantization

   Uniform quantization with deadzone is used to quantize all the wavelet coefficients.

For each subband

b

, a basic quantizer step size D

b

selected by the user and is used to quantize all the is coefficients in that subband.

The choice of the quantizer step size for each subband can be based on visual models, such as the contrast sensitivity function (CSF). This gives higher compression ratio for same visual quality.

Quantization

 Where:  

y

is the input to the quantizer, D

b

is the quantizer step size, 

q

is the resulting quantizer index.

Dequantization

 Where: 

z

is the reconstructed signal value 

r

is the reconstruction bias  R = 0.5 -> midpoint reconstruction (no bias)  R < 0.5 reconstruction is biased to zero (typical is r=3/8)  In JPEG 2000 (Part 1), parameter r is arbitrarily selected by the decoder

Quantization

   Unlike JPEG Baseline, where the resulting quantizer index q is encoded as a single symbol, in JPEG2000 it is encoded one bit at a time, starting from the MSB and proceeding to the LSB.

During this progressive encoding, the quantized wavelet coefficient is called insignificant if the quantizer index q is still zero. Once the first nonzero bit is encoded, the coefficient becomes significant and its sign is encoded.

If the p least significant bits of the quantizer index still remain to be encoded, the reconstructed sample at that stage is identical to the one obtained by using a USQ with deadzone with a step size of D

b 2p

Bit-plane coding quantization

Bit-plane coding quantization example

 Wavelet coefficient: 83. Quantizer step size: 3  Quantizer index: 83/3=27 (00011011)    Dequantized value full: (27+0.5) 3 = 82.5

Dequantized value 6BP:   000110=6 Step size = 3x4=12 Dequantized value = (6+0.5) 12 = 78 Dequantized value 4BP:   0001=1 Step size = 3x16=48 Dequantized value = (1+0.5) 48 = 72

Entropy coding

 Context-based adaptive binary arithmetic coding is used in JPEG2000 to efficiently compress each individual bit plane.

 The binary value of a sample in a block of a bit plane of a subband is coded as a binary symbol with the JBIG2

MQ-Coder

.

Bit coding example

Bit coding example

Bit coding example

Bit coding example

Bit coding example

Bit coding example

Bit coding example

Bit coding example

Resolution SNR quality Full image Visual quality

SNR progressive example

layer resolution

Rate allocation

  Rate allocation is the process that allows to target a specific compression ratio with the best possible quality (MSE, visual or other) for each layer and/or entire codestream. Possible types are:     None: compression ratio is determined solely by the quantization step sizes and image content.

Iterative: quantization step sizes are adjusted according to obtained compression ratio and operation is repeated.

Post-compression: rate allocation is performed after the image data has been coded, in one step.

Others (Lagrangian, scan-based, etc.) Not standardized by JPEG 2000 -> encoder choice.

ROI coding

  

Region of Interest

ROIs.

(ROI) coding allows a non-uniform distribution of quality. The ROI is coded with a higher quality than the background (BG). A higher compression ratio can be achieved with same or higher quality inside

Static

ROIs are defined at encoding time and are suitable for storage, fixed transmission, remote sensing, etc. Commonly referred to as ROI coding.

Dynamic

request.

ROIs are defined interactively by a user in a client/server situation during a progressive transmission. Suitable for telemedicine, PDAs, mobile communications, etc. They can be achieved by the dynamic generation of layers matching the user’s

ROI coding

Error Transmission

    When delivering compressed images across error-prone channels any transmission error can severely affect the decoded image quality. This is specially true since variable length coding is used in the code-block entropy coding and packet heads.

Error types can be random errors, burst error and missing bytes (i.e. network packet loss).

Since each code-block is independently coded an error in a code block’s bitstream will be contained within that code-block. Nevertheless severe distortion can occur in the case of an error.

Packet heads are interdependent and thus fragile.

Example of error resilience

Error management

 Coded image data  Code-block partition  Regular termination of arithmetic coder  Segmentation symbols  Packet heads  Start of packet markers  Packet heads in main / tile codestream header  Partition of packets into precincts

JP2 file format



JP2

is the optional JPEG 2000 file format to encapsulate JPEG 2000 codestreams.

        

Extension: jp2 Allows to embed XML information (e.g., metadata) Alpha channel (e.g., transparency) Accurate color interpretation “True color” and “palette color” supported Intellectual property information Capture and default display resolution File “magic number” File transfer errors (ASCII ftp, 7 bit e-mail, etc.)

Software for JPEG 2000



C (University of Arizona/HP)

Verification model (VM) used in the standard phase of JPEG 2000 

Java (EPFL, Ericsson, Canon)

Part 5 (Reference Software) of JPEG2000 standard ( http://jpeg2000.epfl.ch

) Also a C implementation is publicly accessible

More…

  http://www.jpeg.org

(JPEG official web page) D.S. Taubman; M.W. Marcellin: “JPEG2000: image compression fundamentals, standards and practice”. Kluwer 2002.

 Signal Processing: Image Communication, vol 17(1), pp. 1-144, Jan 2002 (special issue)  Touradj Ebrahimi web page at epfl.ch

MPEG

 Moving Picture Experts Group (MPEG)  Coding of Moving Pictures and Audio  Five Standards:  MPEG-1 - Error free environments  MPEG-2 - Broadcast TV (SD and HD)  MPEG-4 - Object based coding  MPEG-7 - Multimedia description  MPEG-21 - Seven element initiative for multimedia deployment

MPEG 1:1992 (ISO/IEC 11172)

Coding of moving pictures and associated audio for digital storage media at up to about 1,5 Mbit/s

 Part 1 - MPEG-1 Systems - Program Stream  Part 2 - MPEG-1 Video for CD -I  Part 3 - MPEG-1 audio  including Layers I, II, and III(a.k.a. mp3)  Part 4 - Conformance  Part 5 - Technical Report  Directed primarily at error free environments

MPEG 2:1994 (ISO/IEC 13818)

Generic coding of moving pictures and associate audio information

          Part -1 Systems - joint with ITU - 2 ed Dec 2000 Part -2 Video - joint with ITU – 2 ed Dec 2000 Part -3 Audio- forward/backward compatible to MPEG-1(1998) Part -4 Conformance (1998) Part -5 Technical Report (1998) Part -6 DSM CC - Digital Storage Media Cmd and Cntl (1998) Part -7 AAC - Advanced Audio Coding (1997) Part -9 RTI - Real Time Interface (1996) Part -10 Conformance Extensions for DSM-CC (1999) Part -11 IPMP on MPEG-2 Systems  1996 Emmy for technical excellence

MPEG 4:1998 (ISO/IEC 14496)

Coding of audio-visual objects

                  Part 1 Part 2 Part 3 Part 4 Part 5 Part 6 Part 7 Part 8 Part 9 Part 10 Part 11 Part 12 Part 13 Part 14 Part 15 Part 16 Part 17 Part 18 Systems (2001) Visual (2001) Audio (2001) Conformance (2002) Reference Software (2002) DMIF - Delivery Multimedia Integration Framework (2000) Optimized Software (2002) MPEG 4 on IP (2002) Reference Hardware (2003) Advanced Video Coding (AVC) joint with ITU-T (2003) Scene Description and Application Engine ISO Base Media File Format IPMP Extensions MP4 File Format Advanced Video Coding File Format Animation Framework eXtension (AFX) Streaming Text Format Font Compression and Streaming

MPEG 7:2001 (ISO/IEC 15938)

Multimedia content description interface

 Part 1 Systems  Part 2 DDL - Description definition language  Part 3 Visual  Part 4 Audio  Part 5 Multimedia description schemes  Part 6 Reference software  Part 7 Conformance testing  Part 8 Extraction and use of description  Part 9 MPEG-7 Profiles  Part 10 Schema Definition

MPEG 21:2005 (ISO/IEC 21000)

Multimedia framework

             Part -1 Part -2 Part -3 Part -4 Part -5 Part -6 Part -7 Technical Report (2001) Digital item declaration (2002) Digital item identification(2002) IPMP Rights Expression Language Rights Data Dictionary Digital Item Adaptation Part -8 Part -9 Reference Software File Format Part -10 Digital Item Processing Part -11 Evaluation Methods for Persistent Association Part -12 Test Bed for MPEG-21 Resource Delivery Part -14 Conformance Testing