Robust hashing for image authentication using quaternion discrete Fourier transform and log-polar transform

International audienceIn this work, a novel robust image hashing scheme for image authentication is proposed based on the combination of the quaternion discrete Fourier transform (QDFT) with the log-polar transform. QDFT offers a sound way to jointly deal with the three channels of color images. The key features of the present method rely on (i) the computation of a secondary image using a log-polar transform; and (ii) the extraction from this image of low frequency QDFT coefficients' magnitude. The final image hash is generated according to the correlation of these magnitude coefficients and is scrambled by a secret key to enhance the system security. Experiments were conducted in order to analyze and identify the most appropriate parameter values of the proposed method and also to compare its performance to some reference methods in terms of receiver operating characteristics curves. The results show that the proposed scheme offers a good sensitivity to image content alterations and is robust to the common content-preserving operations, and especially to large angle rotation operations