Inverse Halftoning Algorithm Using EdgeBased Lookup Table Approach

1
Inverse Halftoning Algorithm Using Edge-Based
Lookup Table Approach

• Kuo-Liang Chung and Shih-Tung Wu,
• IEEE Transactions on Image Processing,
• Volume 14,  Issue 10,  Oct. 2005, Page(s)1583 -
  1589
• Impact factor 2004 2.011 (27/209 12.9)
• Reporter Jen-Bang Feng (???)

2
Halftone

• Grayscale (color) ? Binary image

3
Halftone Types
Grayscale image

• White Noise
• Clustered-Dot
• Recursive Tessellation
• Blue Noise

White Noise
Clustered-Dat
Recursive Tessellation
Blue Noise
4
Blue Noise(Error-Diffusion)

• Floyd and Steinberg Kernel

R.W. Floyd and L. Steinberg, An adaptive
algorithm for spatial grayscale, Proc. SID, vol.
17, no. 2 , 1976, pp. 75-77.
Origin
Threshold 128
5
Halftone and Inverse Halftone
Grayscale
Halftone
Binary
Inverse Halftone
Recovered Grayscale
6
Look Up Table (LUT)

• Fast, parallel, less computation
• Training the LUT
• Choose the neighborhood or template (a collection
  of pixels) in the prediction.
• Look up the table

Look-up Table (LUT) Method for Inverse
Halftoning M. Mese and P.P. Vaidyanathan IEEE
Transactions on Image Processing, vol. 10, no.
10, Oct. 2001, pp. 1566-1578.
7
Grayscale
Halftone
Grayscale Mapping Table


Training
Pattern
Value
8
Mapping Table
9
LUT Process
For ex.
are in training Sample. Then O 57 if this block
appears
If (p0,p1,,pn-1) exists if not, use white/black
ratio


n 16
Pattern
Value
O denotes the estimated pixel.
10
The Proposed Scheme

• Build LUT
• Build Edge Map by canny
• Classify the edges into sized 4x4
• Regular edges
• Irregular edges
• Train the table with two indices

11
Edge Detection
J. Canny, A computational approach to edge
detection, IEEE Trans. Pattern Anal. Mach.
Intell., vol. PAMI-8, no. 11, Nov. 1986, pp.
679698.
12
Edge Types
Irregular
Regular
13
Grayscale
Halftone
Edge
Grayscale Mapping Table


Training
Pattern
Edge
Value
14
Mapping Table
Edge from LUT
15
The Proposed Table


Pattern
Value




Pattern
Edge index
Value
16
Experimental Results
Recovered by the proposed scheme
Recovered by LUT
17
Experimental Results
PSNR (dB)
18
Conclusions

• An improved look up table inverse halftoning
  scheme
• Considered about edge information

19
Comments

• Other type of edge information
• Post-process of an inverse halftoning