Collusion-Resilient Fingerprinting Using Cropping and Subsampling

Abstract

Valuable images are usually acquired with higher resolution and deeper color depth. Hence, in this paper, we proposed a scheme based on cropping and downsampling to prevent the embedded fingerprints from average attack. We first generate an resized original image for each user with cropping and downsampling. This resized original image belonging to a particular user is then embedded with the fingerprint related to this user. With this method, a coalition of users can still apply average attack on their own fingerprinted images, but the more number of images used, the lower the perceptual quality of the result image will become. The effectiveness of the proposed method is well illustrated by experimental results.