Skip to Main Content
In this paper, a novel fuzzy support vector machine based image watermarking scheme is proposed.Since the application of support vector machine in the process of watermarking technology is only a simple classification of the image. However,the fuzzy support vector machines by selecting the appropriate degree of membership to reflect the different importance of the different sample points. In this article, Firstly, we split the given image into 8 * 8 block, then calculated for each sub-block of the texture features as input vectors to train the support vector machine. The image sub-block is divided into two categories( one category is “-1” represents a weak texture, the other is “+1” represents a strong texture), and we make the strong texture as more important category in this paper. Therefore, given its larger fuzzy membership than the weak texture's. From the relevant theory of watermarking technology we know that a strong local image texture can tolerate more of the watermark information. In order to enhance watermark robustness and embedding more watermark information in the host image, we improve the accuracy of classifying one class(a strong texture). If the data points belonging to the class of strong texture are incorrectly categorized into the weak texture class, the amount of embedded watermarks are reduced and robustness(the ability of resisting attacks) of the image is decreased. Results show that this algorithm has good robustness against common image attacks.