Robust Image Hashing

Background

The hash calculation follows the two steps scheme in Figure 1, consisting in feature extraction and compression. At the end of the first step, the image is reduced at a feature vector called intermediate hash. The intermediate hash is transformed into a robust hash by quantization, binarization and compression. In RIH, the term compression is used purely to designate a significant reduction in the dimensionality of the feature vector. The techniques for obtaining compression in RIH are different from those used in traditional image compression, where the processing chain is quasi reversible in order to allow also the image decompression. In RIH, the reversibility is not necessary.

Figure 1.

RIH general scheme