Digital image hashing using local histogram of Oriented Gradients

Abstract

The ease with which digital images can be manipulated without severe degradation of quality makes it necessary to be able to verify the authenticity of digital images. One way to establish the image authenticity is by computing a hash sequence from an image. This hash sequence must be robust against non content-altering manipulations, but must be able to show if the content of the image has been tampered with. Furthermore, the hash has to have enough differentiating power such that the hash sequences from two different images are not similar. This paper presents an image hashing system based on local Histogram of Oriented Gradients. The system is shown to have good differentiating power, robust against non content-altering manipulations such as filtering and JPEG compression and is sensitive to content-altering attacks.