Geometric attack resistant face watermarking using Zernike moments

Abstract

A robust face watermarking scheme is proposed to provide a solution to copyright protection of semantic content. To address the issue of less robustness to geometric attacks in watermarking scheme, Zernike moments based invariant features are extracted and some selected Zernike moments are proposed to carry a multi-bit watermark. Watermark signal is embedded into face region in the intensity spatial domain after the reconstruction process using the proposed geometric invariants. Experimental results prove the effectiveness of the new scheme, including excellent imperceptibility and robustness to image compression, median filtering and common geometric attacks such as rotation, scaling and translation.