A Neural Network Approach for Visual Cryptography Tai-Wen Yue and Suchen Chiang IEEE 2000
What is Visual Cryptography? Target image Share image2 Share image1
The Procedure Plane-G Plane-S1 (Share 1 ) Plane-H Plane-S2 (Share 2 ) The original taget image Halftone Plain Shadow Image
Image Halftoning Error Aggregated Scan-Line Algorithm e0E0 e0e1E1 e0e1e2E2
The Q ’ tron(Quantum) NN for (2, 2) Plane-G Plane-S1 (Share 1 ) Plane-H Plane-S2 (Share 2 )
Q ’ tron---Q-State Neuron External stimulus Active value Internal stimulus
Q ’ tron---Q-State Neuron Q ’ tron ’ s Input Internal Stimulus External Stimulus Noise
Q ’ tron---Q-State Neuron State Updating Rule:
The Q ’ tron(Quantum) NN Plane-G Plane-S1 (Share 1 ) Plane-H Plane-S2 (Share 2 ) halftoning restoration +
Image Halftoning Energy Function: Sum of graylevel in a 3×3 area for graytone image Sum of halftone in a 3×3 area for halftone image Minimizing the error square corresponds to halftoning -
Image Restoration Energy Function: - (i,j) i j r=3
s1s1 s2s h E2E s1s1 s2s2 h E2E2 FeasibleInfeasible LowHigh Cost Function
Stacking Rule Satisfaction Energy Function: _ 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
Build Plain Shadow Images Energy Function: [g low,g high ] : K 1 =K 2 ≦ g low <g high ≦ min(2K,255) 例如 : [135,235] ; K1=K2=
Total Energy Halftoning Restoration Share Images Cost Function
Conclusions Share images size = target image size. Code book is free.
Future Works Design language to specify an access scheme. Extend to color images