Genetic-based adaptive image watermarking using local image statistics and image features

Abstract

Two novel image watermarking techniques are proposed. The first technique is based mainly on local image statistics and image features, and is referred to as the local image statistics and image features (LISIF) technique. The technique makes the watermark significantly transparent, because two principle image features, edges and smooth regions, are considered in the development of the technique. Moreover, the technique estimates the watermark according to image features and differences between the centre pixel value and the local statistics of each watermarked image block. Thus, the technique does not require information of original images during watermark extraction. However, the performance of the technique depends heavily on five parameters. The second method uses genetic algorithms systematically to seek a set of near-optimal values for the five parameters. Therefore, this method, which is called the genetic-based LISIF (GLISIF) technique, can effectively improve the performance of the LISIF technique. Simulation results are provided in the experiment to claim that the proposed techniques definitely possess not only transparent capability, but also robust and generalized capabilities compared with common image manipulation. Additionally, the results illustrate that the performance of these two techniques is significantly better than that of other proposed methods being considered.