Compensation of geometrical deformations for watermark extraction in the digital cinema application

In this paper, we investigate the restoration of geometrically altered digital images with the aim of recovering an embedded watermark information. More precisely, we focus on the distortion taking place by the camera acquisition of an image. Indeed, in the cinema industry, a large part of early movie piracy comes from copies made in the theatre itself with a camera. The evolution towards digital cinema broadcast enables watermark based fingerprinting protection systems. The first step for fingerprint extraction of a counterfeit material is the compensation of the geometrical deformation inherent to the acquisition process. In order to compensate the deformations, we use a modified 12-parameters bilinear transformation model which closely matches the deformations taking place by an analog acquisition process. The estimation of the parameters can either be global, either vary across regions within the image. Our approach consist in the estimation of the displacement of a number of of pixels via a modified block-matching technique followed by a minimum mean square error optimisation of the parameters on basis of those estimated displacement-vectors. The estimated transformation is applied to the candidate image to get a reconstruction as close as possible to the original image. Classical watermark extraction procedure can follow.