Reversible data hiding for point-sampled geometry

We present a novel reversible data hiding scheme for point-sampled geometry in the spatial domain. To the best of our knowledge, our scheme is the first in the literature for recovering the original point-sampled model using little amount of information. in contrast, other schemes generally need to store an extra large amount of bit size equal to the payload size for reversibility. Our scheme first employs principal component analysis (PCA) for the point-sampled model to produce three principal axes and to construct a PCA-coordinate system. We then translate the coordinates of the original points to the PCA-coordinate system in order to achieve robustness against translation, rotation, and uniform scaling operations. Second, we sort the points' coordinates for each axis to yield intervals which are the embedding positions. Finally, we utilize the left-shift operator to shift the bit of the state value of the interval left by c bits (c >= 1) so that it creates c bits extra storage space for embedding the payload and we store the original state value for achieving reversibility. Experimental results show that our scheme can embed large amounts of data with insignificant visual distortion of the original model. The data capacity in bits achieves nearly 1.5c times the number of points in the models