A fragile watermarking scheme exploiting neural tree for image tamper detection

Abstract

This paper presents a novel watermarking technique based on the Neural tree (NT) classifiers for image authentication and tamper detection. In the proposed technique a neural tree classifier along with multi resolution wavelet analysis is exploited for embedding and extracting the watermark. The NT is trained with input output pairs as the cover image and watermark image, later this trained NT is used to extract the watermark. To insure the authenticity of image a binary sequence is also embedded into the cover image in wavelet domain. The embedding locations of this bit sequence and the trained NT serve as secret key in watermark extraction process. Tamper detection and localization are the main issues in the authentication watermarking schemes. The experimental results show that the proposed technique has good imperceptibility and can detect very minor alterations in the image. © 2012 Springer India Pvt. Ltd.