Robust Image Hashing Based on Hybrid Approach of Scale-Invariant Feature Transform and Local Binary Patterns

Image hash functions find extensive applications in content authentication, database search, and digital forensic. Robust image hash has been widely investigated to authenticate the reliability of images transmitted by a trustless channel. In this paper, we propose a novel image hashing algorithm which is robust to content-reserved and multiple manipulations. To achieve the perceptual robustness and sensitivity, the proposed scheme combines scale-invariant feature transform (SIFT) with local binary pattern (LBP). SIFT extracts plenty of descriptors which are robust to geometric distortion and luminance transformation. LBP generates hash values that contain local information and are sensitive to content manipulation. We further investigate the performance of proposed scheme and other existing algorithms via statistical analysis of recognition rate, and the results show that our method outperforms conventional methods.