Method for watermarking free view video with blind watermark detection

1. A method for watermark embedding in multi views video comprising a plurality of views captured with a plurality of cameras wherein it comprises for each view the steps of:

projecting the view onto a domain invariant to the plurality of views and to intermediate synthesized view, resulting into an invariant vector, wherein said invariant domain comprises a plurality of the samples values along any line parallel to a line passing through optical centers of the cameras, said parallel lines called view invariant dimension, provided that:

i) optical axis of the cameras are parallel,

ii) optical axis are orthogonal to the line passing through optical centers of all cameras,

iii) all cameras share a same focal distance, and

iv) there is no rotation between the cameras;

applying a watermarking algorithm to the invariant vector resulting into a watermarked invariant vector;

subtracting the invariant vector from the watermarked invariant vector obtaining the invariant watermark vector;

distributing the invariant watermark vector over the view invariant dimension resulting into a watermark signal;

merging the watermark signal with the view resulting into a watermarked view.

2. The method according to claim 1 wherein the step of projecting further comprises summing samples values along said invariant dimension of the view.

3. The method according to claim 1 wherein the step of projecting the view further comprises a last step of transforming the invariant vector into a spatial-frequency representation of the invariant vector to which the watermark algorithm is applied.

4. The method according to claim 3 wherein the step of distributing the invariant vector watermark further comprises:

a first step of transforming back the spatial-frequency representation of the invariant watermark vector into an invariant watermark vector.

5. The method according to claim 1 wherein the step of projecting the view further comprises a first step of transforming the view into a spatial-frequency representation of the view.

6. The method according to claim 5 wherein the step of distributing the invariant watermark vector further comprises:

a last step of transforming back the spatial-frequency representation of the watermark to the spatial representation of the watermark.

7. The method according to claim 5 wherein the step of transforming the view into a spatial-frequency representation of the view comprises applying the 1D wavelet transform applied to each column of the view.

8. The method according to claim 1 wherein the step of distributing the invariant watermark vector comprises the duplication of the invariant watermark vector in the invariant dimension of view.

9. The method according to claim 1 wherein the step of merging (E 5 , E 16 ) the watermark signal comprises a step of adding the watermark signal (W) to the view (I).

10. The method according to claim 1 wherein the step of merging the watermark signal further comprises a step of the masking of the watermark signal with a perceptual mask derived from the view.

11. A method for watermark embedding in multi views video comprising a plurality of views captured with a plurality of cameras wherein it comprises for each view the steps of:

generating an invariant watermark vector in a domain invariant to the plurality of views and to intermediate synthesized view, wherein said invariant domain comprises a plurality of the samples values along any line parallel to a line passing through optical centers of the cameras, said parallel lines called view invariant dimension, provided that:

i) optical axis of the cameras are parallel,

ii) optical axis are orthogonal to the line passing through optical centers of all cameras,

iii) all cameras share a same focal distance, and

iv) there is no rotation between the cameras;

distributing the invariant watermark vector over the view invariant dimension resulting into a watermark signal;

merging the watermark signal with the view resulting into a watermarked view.

12. A method for watermark detection in multi views video comprising a rendered view selected among or synthesized from a plurality of views captured with a plurality of cameras wherein it comprises for the rendered view the steps of:

projecting the rendered view onto a domain invariant to the plurality of views and to intermediate synthesized view resulting into an invariant rendered vector, wherein said invariant domain comprises a plurality of the samples values along any line parallel to the line passing through the optical centers of the cameras, said parallel lines called view invariant dimension, provided that:

i) optical axis of the cameras are parallel,

ii) optical axis are orthogonal to the line passing through the optical centers of all cameras,

iii) all cameras share a same focal distance, and

iv) there is no rotation between the cameras;

applying a watermark detection algorithm to the invariant rendered vector in order to assess the presence of a watermark.

13. The method according to claim 12 wherein the step of projecting the rendered view (I′ w ) further comprises a last step of transforming the invariant rendered vector (i′ w ) into a spatial-frequency representation of the invariant rendered vector.