Intelligent reversible watermarking and authentication: Hiding depth map information for 3D cameras

摘要

In this paper, we propose an application for 3D cameras by reversibly hiding the depth map in the corresponding 2D images. The proposed technique is prospective in cameras capable of capturing simultaneously the 2D image and resultant depth map of an object. 3D cameras equipped with self-embedding capability can serve two additional purposes, protection of the captured image and secure transmission of its depth map. The reversible watermarking, in addition to other features, guarantees the lossless recovery of an original image and separation of its depth map when required. For this purpose, a reversible watermarking scheme, based on genetic algorithm (GA), has been proposed which computes suitable threshold for each block of coefficients in the wavelet domain. Thus, a tradeoff is made between watermark imperceptibility and capacity using GA. The threshold map is embedded in addition to the actual payload and thus the proposed approach does not require histogram pre-processing in order to avoid overflow/underflow. The same threshold map has been used for authentication purpose by correlating it with the low-frequency coefficients of the 2D transformed image. Further to exploit the inherent redundancy in the depth map, which is the actual payload in this case, lossless compression has been employed before its embedding. Similarly, besides secret key-based permutation, a cryptographic layer is overlaid on the watermarking layer for security purposes. Experiments conducted on images and depth maps, obtained using a 3D camera and an optical microscopic system, validate the proposed concept.