Neural Networks for Visual Cryptography --- with Examples for Complex Access Schemes Tatung University, Taiwan Presenter: Tai-Wen Yue CAINE-2000
Outline Introduction Neural Network Model --- Q ’ tron NN Q ’ tron NN for Visual Cryptography Experimental Results Conclusions and Feature Works
Introduction
What is visual cryptography? ( n, k )-scheme: k out of n Decompose a secret image into a set of n shadow images called shares. A share carries meaningless information. Stacking k or more shares, printed on transparencies, reveals the secrete. Decrypting using eyes
Example Target image Share image2 Share image1
Applications Key Management Message Concealment Authorization Authentication Identification Entertainment
Access Schemes Shares Stacking all shares Stacking two shares (2, 2)(3, 2)Full
Traditional Approach Using codebooks An Example codebook: (2, 2) PixelProbability Shares #1 #2 Superposition of the two shares White Pixels Black Pixels
Our Approaches No codebook required Inputs are gray images Target Image(s) Share Images Outputs are halftone images that mimic the corresponding gray images Applicable to complex access schemes
Neural Network Model Q ’ tron NN
Q ’ tron Active value Weighted and multilevelled Each Q ’ tron represents a quantity --- a i Q i
Q ’ tron Active value Internal stimulus Input due to Q ’ trons ’ Interactions T ii usually is nonzero and negative
Q ’ tron Active value Internal stimulus External stimulus External input serves as bias
Q ’ tron Active value Internal stimulus External stimulus Escape local-minima Persistent noise --- no holiday
Q ’ tron External stimulus Active value Internal stimulus
State Transition Rule Q ’ tron ’ s Input Internal Stimulus External Stimulus Noise Free
State Transition Rule State Updating Rule: Running Asynchronously
Q ’ tron NN vs. Hopfield NN Running Asynchronously Noise Free T ii =0 q i =2 Noise Free T ii =0 q i =2 Q ’ tron NN = Hopfiled NN
Energy Function Interaction Among Q’trons Interaction with External Stimuli Constant Monotonically Nonincreasing
Problem Solving Using a Q ’ tron NN A given problem A optimization problem Reformulation Cost Function Energy Function Build Q’tron NN Mapping
Operation modes External stimulus Active value Internal stimulus Clamp-mode
Operation modes External stimulus Active value Internal stimulus free-mode
Why operation modes? Unstable Stable
Why operation modes? Clamped Free
Why operation modes? Clamped Free
Q ’ tron NN for Visual Cryptography Highlight the main concept by (2, 2)
The Q ’ tron NN for (2, 2) Plane-G Plane-S1 (Share 1 ) Plane-H Plane-S2 (Share 2 )
The Q ’ tron NN for (2, 2) Plane-G Plane-S1 (Share 1 ) Plane-H Plane-S2 (Share 2 ) Target Image Clamped
The Q ’ tron NN for (2, 2) Plane-G Plane-S1 (Share 1 ) Plane-H Plane-S2 (Share 2 ) Target Image Clamped Share 1 + Share 2 Share 1
The Q ’ tron NN for (2, 2) Plane-G Plane-S1 (Share 1 ) Plane-H Plane-S2 (Share 2 ) Target Image Clamped Share 1 + Share 2 Share 1 Halftoning Stacking Rule
Halftoning + Stacking Rules Halftoning Gray Images Binary Images Gray Images: Target and Shares Stacking Rules Fulfill the Access Scheme
Halftoning Graytone Image Halftone Image Halftoning How? To make the average luminances of each cell-pair as close as possible.
Halftoning Gray Image Halftone Image Halftoning May have many solutions
Stacking Rules Gray Image Halftone Image Halftoning Share Images Stacking Rule One or more pixels black Black
Energy function --- Halftoning A 3 3 halftone cell A 3 3 graytone cell The luminance difference (squared error)
Stacking Rules (The magic) s1s1 s2s h E2E s1s1 s2s2 h E2E2 Feasible Infeasible + = s1s1 s2s2 h
Stacking Rules (The magic) s1s1 s2s h E2E s1s1 s2s2 h E2E2 Feasible Infeasible + = s1s1 s2s2 h LowHigh
Energy function --- Stacking Rules Minimizing this term tends to satisfy the stacking rules
Share Image Assignment For simplicity, shares are plain images S1 S2 Mean Gray level K 1 K2K2 Result
Energy Function--- Share Image Assignment
Total Energy Halftoning Stacking Rules Stacking Rules Share Images Share Images
Q ’ tron NN Construction Mapping
Experimental Results
Histogram Reallocation Needed + + Histogram Reallocation
The Procedure Plane-G Plane-S1 (Share 1 ) Plane-H Plane-S2 (Share 2 ) The original taget image
The Procedure Plane-G Plane-S1 (Share 1 ) Plane-H Plane-S2 (Share 2 ) The original taget image Histogram Reallocation Clamped
The Procedure Plane-G Plane-S1 (Share 1 ) Plane-H Plane-S2 (Share 2 ) The original taget image Histogram Reallocation Clamped Free
Experimental Result --- (2, 2) Share 1Share 2 Target Image Share 1 + Share 2
Generalized Access Scheme Experimental Results
Full Access Scheme Shares 朝辭白帝彩雲間 朝 辭 白 帝彩雲 間 Shares
Full Access Scheme Shares 朝辭白帝彩雲間 朝 辭 白 帝彩雲 間 Shares Theoretically, unrealizable. We did it in practical sense.
Full Access Scheme Shares S1S2S3 S1+S2S1+S3S2+S3S1+S2+S3
Access Scheme with Forbidden Subset(s) 人之初性本善 人 之 初 性本 X 善 Theoretically, realizable. Shares
Access Scheme with Forbidden Subset(s) S1S2S3 S1+S2S1+S3S2+S3S1+S2+S3
Access Scheme for Access Control S1S2S3 S4S1+S4S2+S4S3+S4
Target and Shares are Gray Images S1 Armored knight
Target and Shares are Gray Images S2 Man
Target and Shares are Gray Images S1 + S2 Armored Knight + Man = Lina
Conclusions and Future works
Conclusions How? NNs for visual cryptography No codebook. Uniform math for access schemes. Target images and share images are graylevelled ones Share image size = Target image size
Future Works Design language to specify an access scheme. Auto generation of the Q ’ tron NNs Histogram Reallocation is a nontrivial task. Extend to color images