Pixel Grouping Based Image Hashing for DIBR 3D Image

Most of the traditional 2D image hashing schemes do not take into account the change of viewpoint to construct the hash vector, resulting in the classification accuracy rate is unsatisfactory when applied in identification for Depth-image-based rendering (DBIR) 3D image. In this work, pixel grouping according to histogram shape and Nonnegative matrix factorization (NMF) is applied to design DIBR 3D image hashing with better robustness resist to geometric distortions and higher classification accuracy rate for virtual images identification. Experiments show that the proposed hashing is robust to common signal and geometric distortion attacks, such as additive noise, blurring, JPEG compression, scaling and rotation. When compared with the state-of-art schemes for traditional 2D image hashing, the proposed hashing provides better performances under above distortion attacks when considering the virtual images identification.