Lossy Compression of Stochastic Halftones with JBIG2

1
Lossy Compression ofStochastic Halftones with
JBIG2

• Magesh Valliappan and Brian L. Evans
• Embedded Signal Processing Laboratory
• The University of Texas at Austin
• http//signal.ece.utexas.edu/

Dave A. D. Tompkins and Faouzi Kossentini Signal
Processing and Multimedia Group University of
British Columbia http//spmg.ece.ubc.ca
2
Introduction

• Digital halftoning
• Continuous tone to bi-level
• Ordered dithered halftones
• Periodic mask of thresholds
• Stochastic halftones
• Shape quantization noise intohigh frequencies

3
Joint Bi-Level Experts Group

• JBIG2 Standard
• Document printing, faxing, scanning, storage
• Lossy and lossless coding
• Models for text, halftone, and generic regions
• Lossy JBIG2 compression of halftones
• Preserve local average gray level not halftone
• Spatially periodic descreening
• High compression of ordered dither halftones

4
Motivation

• Improve JBIG2 performance on stochastic halftones

Existing JBIG-26.1 1
Proposed Method 6.6 1
5
Lossy Compression of Halftones
Generate (M21) patterns ofsize M x M from a
clustereddot threshold mask
Construct Pattern Dictionary
Lossless Encoder
JBIG2 bitstream
Halftone
Compute Indices into Dictionary
Count black dots in each M x M block of
input Range of indices 0... M21
6
Proposed Method

• modified multilevel Floyd Steinberg error
  diffusion
• optionally reduce N
• L sharpening factor
• one multiply/add

• 3 x 3 lowpass
• zeros at Nyquist
• removes noise
• artifacts
• power-of-two coefficients

Prefilter
Decimator
Quantizer
Lossless Encoder
JBIG2 bitstream
Halftone
17
16 M2 1
graylevels
N
2

• M x M lowpass averaging filter
• downsample by M x M

Symbol Dictionary

• N patterns
• size M x M
• may be angled
• clustered dot

7
Quality Metrics

• Model degradation as linear filter plus noise
• Decouple and quantify linear and additive effects
• Contrast sensitivity function (CSF) C(?1, ?2)
• Linear shift-invariant model of human visual
  system
• Weighting of distortion measures in frequency
  domain

8
Quality Metrics

• Estimate linear model by Wiener filter
• Weighted Signal to Noise Ratio (WSNR)
• Weight noise D(u , v) by CSF C(u , v)

• Linear Distortion Measure
• Weight distortion by input spectrum X(u , v) and
  CSF C(u , v)

9
Results
High Quality Ratio 6.6 1 WSNR 18.7 dB LDM 0.116
High Compression Ratio 9.9 1 WSNR 14.0
dB LDM 0.158
512 x 512 Floyd Steinberg halftoneof barbara
image
10
Results
Results for 512 512 Floyd Steinberg halftone
11
Rate Distortion Curve - LDM
12
Rate Distortion Curve - WSNR
13
Conclusions

• JBIG2 encoding of stochastic halftones
• Reduce noise and artifacts
• Achieve higher compression ratios
• Require low computational complexity
• Rate distortion tradeoffs of free parameters
• Quality metrics consistent with visual quality

14
Backup Slides
15
Introduction

• Compression of halftones
• Document printing, faxing, scanning, storage
• 4" 4" 600dpi halftone (1bpp), 720,000 bytes
• Lossy compression
• Details invisible to human eye
• Only underlying grayscale image important
• Perceptually near lossless at high compression

16
Standards

• Group 3 fax standard (1980)
• 1-D data model
• lossless run length encoding and Huffman encoding
• Group 4 fax standard (1985)
• 2-D data model
• Joint bi-level experts group 1 (1995)
• 2-D lossless context based arithmetic coding