An improvement of RGB color image watermarking technique using ISB stream bit and Hadamard matrix

Abstract

In the past half century, the advancement of internet technology has been rapid and widespread. The innovation provides an efficient platform for human communication and other digital applications. Nowadays, everyone can easily access, copy, modify and distribute digital contents for personal or commercial gains. Therefore, a good copyright protection is required to discourage the illicit activities. On way is to watermark the assets by embedding an owner's identity which could later on be used for authentication. Thus far, many watermarking techniques have been proposed which focus on improving three standard measures, visual quality or imperceptibility, robustness and capacity. Although their performances are encouraging, there are still plenty of rooms for improvements. Thus, this study proposes a new watermarking technique using Least Significant Bit (LSB) insertion approach coupled with Hadamard matrix. The technique involves four main stages: Firstly, the cover image is decomposed into three separate channels, Red, Green and Blue. Secondly, the Blue channel is chosen and converted into an eight bit stream. Thirdly, the second least signification bit is selected from the bit stream for embedding. In order to increase the imperceptibility a Hadamard matrix is used to find the best pixels of the cover image for the embedding task. Experimental results on standard dataset have revealed that average PSNR value is greater than 58db, which indicates the watermarked image is visually identical to its original. However, the proposed technique suffers from Gaussian and Poisson noise attacks.