This paper continues our researches on storage and bandwidth reduction for stereo images by using reversible watermarking. By embedding into one frame of the stereo pair the information needed to recover the other frame, the transmission/storage requirements are halved. Furthermore, the content of the image remains available and one out of the two images is exactly recovered. The quality of the other frame depends on two features: the embedding bit-rate of the watermarking and the size of the information needed to be embedded. This paper focuses on the second feature. Instead of a simple residual between the two frames, a disparity compensation scheme is used. The advantage is twofold. First, the quality of the recovered frame is improved. Second, at detection, the disparity frame is immediately available for 3D computation. Experimental results on standard test images are provided.

The distortion introduced by reversible watermarking depends on the embedding bit-rate. This paper proposes a fine control of the embedding bit-rate for low capacity histogram shifting reversible watermarking. The basic idea of our approach is to split the prediction error histogram into several histograms and to ensure the fine tuning of the bit-rate by selecting the appropriate bins for each histogram. The splitting is performed as a function of the prediction context. The bins (two for each histogram, one for the right side, another for the left side of the histogram) are selected by linear programming in order to minimize the distortion introduced by the watermarking. The proposed scheme outperforms in terms of embedding distortion the prior state of the art.

By embedding into the three color channels of a stereo frame the information needed to recover the other frame, the transmission/storage requirements when operating with stereo images are halved. A disparity compensation scheme on each color plane is used. Due to the reversibility of the watermarking, the embedded frame is recovered exactly. The quality of the recovered frame depends on the embedding bit-rate of the reversible watermarking method and of the information size needed to be embedded. The experimental results obtained so far on high quality color images are very good.

This paper investigates the block implementation of median edge detector (MED) based prediction error expansion reversible watermarking. Image is divided in overlapping blocks. A block is considered for data embedding only if all its pixels can be embedded. By embedding on blocks, the size of the location map is reduced. The block map implementation provides considerably better results than the classical location map implementation and rather similar to the histogram shifting ones. Experiments for different block sizes are provided.