Agustín Moreno Cañadas

Nelly Paola Palma Vanegas

Digital halftoning

Gray Scale Image Halftoned Image

FUSION

DHSPT

Darrell, Fisher, Viola and Freeman explain this effect as human capacity to identify a sound from several ones especially when there is different kind of noise. The cocktail party effect can be analyzed as two related problems: Recognition: how do humans segregate speech sounds, and is it possible build a machine to do the task? What cues in the original are important for separating one voice from other conversations and background noise? Synthesis: can be used to enhance a listener's ability to separate one voice from another in an interactive speech system.

Cocktail Party image

Secret Image

Shares

Overlapped Image

Pixel Expansion

Extended Visual Criptography

Secret Image

Shares

Share 1 Share 2

Share 3

Share 1 + Share 2 Share 2 + Share 3

Share 1 + Share 2 + Share 3 Share 1 + Share 3

Fiducial Points

Face Bunch Graph (FBG)

Jet

Elastic Bunch Graph Matching

Encoding Process

In order to encode a fixed set of gray-level images x1, x2, …, xt we proceed as follows: a. Fix a distinguished set X = {X1, X2, …, XD} subset of P of

participants.

b. Define a system of maximal host images J1, J2, …, JS.

c. Define a secret elastic bunch G(IJih ) to all image IJi

h . These graphs may be obtained by stacking special subsets of transparencies.

d. Finally, split up Γ’ into D disjoint subsets giving one of these subsets to each distinguished participant, in such a way that a permutation π, of Γ’ may be constructed.

Decoding Process We proceed as follows in order to construct the transparencies for a given binary image xi’ subset of Jh: a. Define π-1 in order to construct Γ’.

b. Define G(xi).

c. If u is a pixel in image e(Xxi’) with location l(u) and pixel value p(u) then

the corresponding pixel values in transparencies σ(u) satisfy the condition:

σ(u) = V p(ut) = 0, where t belongs to T(e(Xxi’)), the minimum set of

transparencies (or threshold) required to decoding image xi’ , ut is the pixel in the transparency t in T(e(Xxi’)) such that l(ut) = l(u). σ(u) = 1, if u belongs to Γ’ - e(Xxi’).

d. In this step, in order to eliminate all spurious pixels.

Due that, there are D! possible arrangements for Γ’, we can see that the inclusion of the distinguished set provides additional security to the scheme.

Γ’ Elastic Bunch

Decoding Example

Text Plain

Another Example

The contribution of this work is an approximation to GEVCS where the decoding process can be viewed as a cocktail party effect. It involves some biometric techniques which allow us to give a bound for the corresponding pixel expansion process. In particular, using such biometric techniques, in the decoding process, allow us to obtain a new method to break DHSPT .