SMART: Saliency Map, Moments and Texture Features for Robust Image Authentication

Image Authentication is the process of providing proof to a person or system that the images are indeed that claims to be. Image hashing is a technique used for image authentication and is very popular and ensures remarkable results. Global and Local feature extraction is enhanced for generating a hash that is sensitive to image content. Normal manipulations like JPEG coding, rotation, cropping, additive noise, gamma correction and scaling are robust enough so that the image content information can be recognized as authentic even if the image intensity is varied by these operations. The concern regarding content preserved hashing is that the salient regions need to be detected accurately so that local feature extraction can be applied in to the region of interest. A hybrid approach consisting of frequency prior, center prior and location prior along with watershed for segmentation is used for extracting the saliency regions in the image more efficiently with increasing localization capability in less hash length.