Binocular suppression based visual masking model for stereo image watermarking

Digital watermarking algorithms for stereo image and video have been simple extensions of monoscopic watermarking methods to two views, and have not considered the aspects of the human visual system which are particular to stereo perception. One key aspect is binocular suppression, which says that high quality image data (for example with sharp edges) presented to one eye can suppress low quality information (for example, blurriness) which is simultaneously presented to the other eye, so that the combined subjective visual effect is one of high quality. We exploit this phenomenon by inserting watermark bits into different portions of the left and right image views, so that high quality in a region on one view suppresses the visibility of the watermark in the corresponding region of the other view. Tests with human observers confirm the visual superiority of the approach.