Download presentation
Presentation is loading. Please wait.
Published byBenedito Paranhos Modified over 5 years ago
1
High-Capacity Data Hiding in Halftone Images Using Minimal-Error Bit Searching and Least-Mean Square Filter Author: Soo-Chang Pei and Jing-Ming Guo Source: IEEE Transactions on Image Processing, Vol. 15, No. 6, pp , June 2006 Date: 26/10/2006 Speaker: Ming-Ni Wu
2
Outline Introduction Error Diffusion PSNR for Halftone Images
Data Hiding with Minimal-Error Bit Search (MEBS) Standard MEBS Modified MEBS Self-Decoding Mode Dot-Diffusion Self-Decoding Mode Color Halftone Image Self-Deconding Mode Robust Watermarking Conclusions
3
Introduction Error Diffusion (EDF) A gray level image A halftone image
4
Introduction to distribute the watermark into several EDF images (secret sharing) Error Diffusion and Data Hiding gray level images halftone images a bi-level watermark
5
Error diffusion (EDF) 255 80 100 95 200 150 130 72 137 160 110 63 121 143 90 54 TH=128, x=80 80<128=>x=0 Error=80-0=80 100+80*(7/16)=135 200+80*(3/16)=215 150+80*(5/16)=175 130+80*(1/16)=135 255 135 95 215 175 72 137 160 110 63 121 143 90 54 a grey level image x 7/16 3/16 5/16 1/16 x 7/48 6/48 3/48 5/48 1/48 Steinberg kernel Jarvis kernel
6
Error diffusion (experiments)
Original error-diffused halftone images 512 512 PSNR=30.84 PSNR=27.15 PSNR=30.9
7
PSNR for Halftone Images
Q 1 150 80 100 95 200 130 72 137 160 110 63 121 143 90 54 0.05 0.10 0.40 P Filter W Halftone image B Original gray level image 0.59 0.31 0.39 0.37 0.74 0.51 0.28 0.54 0.63 0.43 0.25 0.47 0.56 0.35 0.21 0.05*1+0.1*0+0.05*1+0.1*1+0.4*1+0.1*0+0.05*0+0.1*1+0.05*0 =0.8 Normalized image X
8
PSNR for Halftone Images
-0.002 -0.001 0.006 -0.010 -0.03 -0.04 -0.004 0.011 0.033 0.042 0.022 -0.006 0.035 0.081 0.099 0.063 0.016 0.040 0.097 0.123 0.027 0.002 -0.003 0.013 0.053 0.073 0.049 0.014 -0.007 0.012 -0.005 7*7 filter
9
Standard MEBS X Y secret image 255 80 100 95 200 150 130 72 137 160
110 63 121 143 90 54 x2=80,80<128 =>x2’=0 y2=130,130>128=>y2’=1 Case 1: x2’=1, Error=80-255=-175 y2’=1, Error= =-125 Total error =|-175|+|-125|=300 Case 2: x2’=0, Error=80-0=80 y2’=0, Error=130-0=130 =|80|+|130|= selected secret bit 1 Gray code X 130 120 100 70 145 85 50 80 87 75 63 66 Y 1 secret image
10
Standard MEBS (Experiments)
watermarked images PSNR= PSNR=25.45 watermark capacity:50%
11
Standard MEBS (Experiments)
watermarked images PSNR=28.9 PSNR=26.9 PSNR=28.95 watermark capacity:33%
12
Modified MEBS X Y secret image secret bit 1 Gray code 255 80 100 95
ERRth=128 x2=80,80<128 =>x2’= 0 y2=130,130>128=>y2’=1 Case 1: x2’=1, Error=80-255=-175=>-128 y2’=1, Error= =-125 Total error =|-128|+|-125|=253 Case 2: x2’=0, Error=80-0=80 y2’=0, Error=130-0=130=>128 =|80|+|128|= selected secret bit 1 Gray code 255 80 100 95 200 150 130 72 137 160 110 63 121 143 90 54 X 130 120 100 70 145 85 50 80 87 75 63 66 Y 1 secret image
13
Modified MEBS The average ERRth value in TABLE II is
14
Modified MEBS (Experiments)
watermarked images PSNR= PSNR=26.78 watermark capacity:50%
15
Dot-diffusion Self-Decoding Mode
Dot Diffusion and Data Hiding gray level images halftone images a bi-level watermark
16
Dot-diffusion Self-Decoding Mode
16 60+70*(1/7)=70 50+70*(2/7)=70 80+70*(1/7)=90 90+70*(2/7)=110 90+70*(1/7)=100 16 16 16 … … 26 19 49 39 47 78 62 63 94 255 60 70 50 80 90 255 70 90 110 100 … … orthogonal factor=2 diagonal factor=1
17
Dot-diffusion Self-Decoding Mode (experiments)
watermarked, 480*480, PSNR=25.39 capacity=50% watermark,240*480 Watermarked, 480*480, PSNR=28.84 capacity=33.3% watermark,160*480
18
Color Halftone Image Self-Deconding Mode
B B Error Diffusion and Data Hiding a color image a color halftone image a bi-level watermark
19
Color Halftone Image Self-Deconding Mode (experiments)
RGB CMYK
20
Robust Watermarking 512 64 64 512 8 Each block can embed the same information bit. We can use the majority voting to recover the damaged portion. 8
21
Robust Watermarking (experiments)
Tampering attack 92.65%
22
Robust Watermarking (experiments)
Cropping attack 10% % % % 99.54% % % %
23
Robust Watermarking (experiments)
Rotated attack (a)10° (i)95.87% (b)15° (j)94.12% (c)20° (k)92.55% (d)25° (l)91.24% (e)30° (m)90.65% (f)35° (n)90.14% (f)40° (o)89.94% (h)45° (p)89.92%
24
Conclusions A high-capacity data-hiding method in halftone image using the MEBS technique is proposed. The embedded halftone images are robust enough to survive under various attacks.
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.