A robust depth-image-based-rendering 3D image hashing scheme based on histogram shape

Using the approximate invariance of the image histogram shape to geometric distortions especially the DIBR process, we propose a hashing scheme for DIBR 3D images by selecting several pixel groups to construct the image histogram shape, and computing the ratios of pixel numbers in different groups to generate the final hash sequence. Meanwhile, we use an N-times searching method to improve the robustness of proposed hashing scheme. Experimental results show that the proposed method can achieve better identification performances under geometric attacks such as rotation attacks, and provide comparable performances under classical distortions such as additive noise, blurring, and compression. Furthermore, this method can ensure that the generated virtual images could be identified with the same content as the corresponding center image when we adjust the baseline distance.