Maximum Likelihood Decoder For Image Watermarking Techniques

Abstract

The paper extends a general approach for extracting the watermark using a mathematical technique for watermarking using Discrete Wavelet Transform (DWT). Though the watermark may be inserted using quantization or a combination of other transforms, yet the successful recovery of watermarks is a difficult task. It is indeed a challenge for the decoder to recover the watermark when various attacks are applied upon a watermarked image. The corresponding scheme of watermark decoding uses the principle of Maximum Likelihood (ML) estimation and is based upon the statistical modeling of parameters of Gaussian distribution. The watermarking experiments utilizing ML decoder exhibit improved performances over the conventional approaches of watermarking employing correlation detectors. The robustness of such techniques is verified through the values of Bit Error Rate (%).