Robust Line Based Image Fingerprinting Algorithm

Image fingerprinting is an important technology that is used by many search engines and multimedia systems for indexing, copy detection and retrieving images. However, most fingerprinting algorithms are point based algorithms. This restriction causes an impact on algorithm robustness against image manipulations. In addition, hashing algorithms that are based on dimensionality reduction have the problem of revealing the original image or low quality version of it in the fingerprint. This paper presents a robust fingerprinting algorithm that is based on line segments encoding. This algorithm calculates fingerprints based on the relations between image line segments instead of image points. The motivation is generating fast robust fingerprints that are resilient to scaling, rotation, translation, color effects, added noise and image cropping. The generated fingerprint should not reveal any part of the original image and could not be reverse engineered to recover the original image.