Post on 08-Jan-2016
description
A Quick Illustration of JPEG 2000
Presented by
Kim-Huei LowChun Data Fok
Overview Introduction Approach Illustration
Annex B-H Comparison with JPEG
Conclusion References Questions
Figure:Picture of Data using
JPEG (75% compression ratio,
15KB)
Figure:Picture of Kim
using J2K (0.5bpp, 3.8KB)
Introduction JPEG 2000
Drafted by the international JPEG (Joint Bi-level Image Experts Group) and JBIG (Joint Photographic Experts Group) groups.
Replaces traditional JPEG. Focuses on hardware implementation.
Our goal Present a simplified version of the standard. Give new users a grasp of JPEG 2000.
Approach Follow the same order
as the standard. Explain the
background. Illustrate each feature. Discuss its
applications. List the pros and cons. Will skip Annex A, C
and D. Feature wise, it’s not
important.
Figure: 0.25bpp J2K Image (11KB); Raw Image’s Size is 1MB
Illustration: Annex B Tile division
Large images can be broken down into smaller pieces, called tiles.
Tiles are processed independently
Figure: Original DWT Figure: Precinct Selection
Figure: Sub-band Selection Figure: Code-block Selection
Illustration: Annex B Progression Order
Layer or Resolution Progressive
Figure: 1bpp, 0.5bpp, 0.05bpp and 0.01bpp J2K Image with Layer or Resolution Progression.
Illustration: Annex B Progression Order
Component Progressive
Figure: 1bpp, 0.5bpp, 0.1bpp and 0.01bpp J2K Image with Component Progression.
Illustration: Annex B Progression Order
Position Progressive
Figure: 1bpp, 0.5bpp and 0.1bpp J2K Image with Position Progression.
Illustration: Annex E Quantization
Reversible vs Irreversible Target bit rate=0.5 bpp Step size=1
Figure: Reversible Quantization (16KB)
Figure: Irreversible Explicit Quantization (868B)
Figure: Irreversible Implicit Quantization (787B)
Illustration: Annex E Irreversible Explicit Quantization
Target bit rate=0.5 bpp Different step size
Figure: Step Size 1 (868B) Figure: Step Size 0.1 (11.9KB) Figure: Step Size 0.0078 (16.3KB)
Illustration: Annex E Irreversible Implicit Quantization
Target bit rate=0.5 bpp Different step size
Figure: Step Size 1 (787B) Figure: Step Size 0.1 (11.7KB) Figure: Step Size 0.0078 (16.3KB)
Illustration: Annex F Discrete Wavelet Transformation (DWT)
Reversible = 5x3 filter (lossless compression) Irreversible = 9x7 filter (efficient lossy compression)
Illustration: Annex F Lossless vs Lossy DWT
Different decomposition level Higher decomposition levels – higher overhead
Figure: Lossless, NL=14 (275KB)
Figure: Lossy, NL=14 (99KB)
Figure: Lossless, NL=3 (274KB)
Figure: Lossy, NL=3 (98KB)
Illustration: Annex F Discard of high frequency sub-bands
High compression, smaller file size Same quality Amortize decomposition level overhead Optimal/Ideal: Encode up to the last visually
distinguishable low frequency sub-band
Figure: NL=3, 8.3636bpp (274KB)
Figure: NL=14, 0.9948bpp (32.6KB)
Illustration: Annex G DC Level Shifting
Similar to JPEG New pixel value = Pixel value - 128
Component Decorrelating Transformation Reversible vs Irreversible
Figure: Raw Image; 0.035bpp J2K Image with RCT; 0.035bpp J2K Image with ICT
Illustration: Annex H Region of Interest (ROI) Encoding
Efficient use of bit rate If bit rate is too low, encoding without ROI may
look better overall
Figure: Raw Image; 0.07bpp J2K Image with ROI; 0.07bpp J2K Image without ROI
Illustration: Comparison of JPEG 2000 with JPEG 10 test images, 50+ compression ratios PSNR vs File Size
xy
yxfyxg
MNPSNR
|),(),(|
255log10 10
PSNR vs File Size
0
5
10
15
20
25
0 10000 20000 30000 40000 50000 60000 70000 80000
File Size (bytes)
PS
NR
(d
B)
JPEG
JPEG 2000
Figure: PSNR Curve
Illustration: Comparison of JPEG 2000 with JPEG Much smaller files Much better quality
Figure: 0.08bpp J2K Image (8KB); 0.1563bpp JPEG Image (16KB);
Conclusion Excellent compression rate Fully exploits the advantage of DWT Capable of handling extremely large
images Lots of user-selectable features Efficient for hardware implementation Most advanced image compression
standard Implication of MPEG 2000?
References International Telecommunication Union (ITU),
International Organization for Standardization (ISO), “JPEG 2000 Implementation in Java,” http://jpeg2000.epfl.ch, October 16th, 2003.
ISO/IEC JTCI/SC29 WGI, JPEG 2000 Editor Martin Boliek, Charilaos Christopoulos, Eric Majani, “JPEG 2000 Image Coding System,” http://www.jpeg.org/CDs15444.html, March 16th, 2000.