Transcript 下載/瀏覽
A novel secret image sharing scheme for true-color images with size constraint Source: Information Sciences, Vol. 179, No. 19, Sep.2009,pp.3247-3254. Author: Tsai Du-Shiau, Horng Gwoboa, Chen Tzung-Her, Huang Yao-Te Article history: Received 2 August 2008 Received in revised form 21 May 2009 Accepted 29 May 2009 Advisor: Chih-Hung Lin Speaker: 趙家寬 1 Outline Introduction Review of related works The proposed scheme Experimental results Discussion Conclusions 2 Introduction(1/3) Secret Image Sharing (SIS) 1. 2. Distributing partial information, called shares Two types of SIS Lossless Lossy Function f () X i ' f ( X i ) ; and X i f 1 ( f ( X i )) . 3 Introduction(2/3) Three main categories of SIS VC-based(視覺密碼學) 2. VVSS-based (變異視覺式機密分享) 3. IM-based(插值方法) 1. 4 Introduction(3/3) Four guidelines to propose a new SIS scheme 1. 2. 3. 4. Support 24-bit true-color image Size of each share is slightly increased Camouflage images for hiding secrets should be natural Visual perception and image properties 5 Review of related works(1/7) 1. VC-based secret image sharing schemes (Proposed by Naor and Shamir) 6 Review of related works(2/7) Characteristics 1. 2. 3. 4. Perfectly secure Decode secrets by visual systems with no cryptography knowledge and without performing cryptographic computation Transparencies are meaningless Size of each transparency is much larger than that of secret images 7 Review of related works(3/7) VVSS-based secret image sharing schemes (Proposed by Chang et al.) 2. • • • Give three n x m pixels image - C-color secret image S and two randomly chosen cover images O1 and O2 Colors of S is stored in the Color Index Table (CIT) Each pixel in O1 and O2 is expanded into minimized M t t subpixels to obtain two camouflage images, O1’ and O2’ 8 Review of related works(4/7) M satisfies : (3) (4) 9 Review of related works(5/7) Bit-level based SIS scheme (Proposed by Lukac and Plataniotis) 10 Review of related works(6/7) 4. PCA and FFN neural networks Principal component analysis (PCA) Color image compression scheme with hybrid neural network consisting of Sanger’s algorithm (Proposed by Clausen and Wechsler) 11 Review of related works(7/7) The neural network begins with initial A and learns/updates weights aij according to A feed forward neural network trained with back propagation also begins with initial B and learns/updates weights bij according to 12 The proposed scheme(1/4) 1) Initial phase Two global weighted matrices A and B 13 The proposed scheme(2/4) Encoding phase 2) • • Input: a true-color secret image X Output: n camouflage images (11) 14 The proposed scheme(3/4) (12) (13) 15 The proposed scheme(4/4) Decoding phase 3. • • Input: n camouflage images, global weighted matrix B, seed s Output: true-color secret image X’ 16 Experimental results 17 Discussion(1/2) 1) 2) 3) 4) 5) 6) Support the 24-bit true-color secret image Size constraint on shares Quality improvement Efficiency of bandwidth and storage space Security enhancement Low computational power 18 Discussion(2/2) 19