This paper proposes a Robust Image Hashing Based on Multi-view Feature Representation and Tensor Decomposition. The algorithm effectively utilizes complementarity and independence between multiple views, and combines the characteristics of multiple views to construct a binary robust hash code. The input image is first prepossessed by bilinear interpolation and mean filtering. Next, the multi-view features, i.e. structure features, edge features and color features are extracted. The correlation

more »

... etween the multiple views is explored to construct a high-order tensor. Then, the popular algorithm Tucker Decomposition (TD) is applied to decompose the tensor into core tensor and factor matrices. Finally, a robust hash code is constructed by using factor matrices. The experimental results of this paper show that the algorithm can resist a large number of content preservation operations and has strong discrimination. The ROC curve obviously demonstrates the performance advantage of the proposed algorithm compared with other algorithms.