Content-based image authentication using sparse features and SOM trajectories
Alberto Peinado, Andrés Ortíz, Guillermo Cotrina · Logic Journal of IGPL · 2015
In this article, we propose a hash function designed to implement a content-based image authentication mechanism using Self-Organizing Map (SOM) trajectories and sparse features extracted from images. The features computed are used as inputs to the SOM which computes a number of prototypes. After that, the prototypes, corresponding to each image block, define a trajectory in the SOM space, which is used to define the hash. Moreover, the image transformations can be identified as they impose changes in hash. Unlike previous works, which use local histogram features and statistical descriptors, we found that trajectories computed using sparse features present a more stable behaviour with small SOM maps.