A watermarking scheme using image object recognition

Abstract

Most existing watermarking techniques embed watermarks into the entire image without taking the image object region into account. For one kind of image with salient object feature, such as portraits, the object region is more important than other parts. In this paper, we propose a watermarking scheme using image object region, in which we exploit DWT to decompose the image and cluster the high-frequency wavelet coefficients to obtain the embedding domain of the watermarks. In this domain, we embed watermarks based on quantization modulation. Experiments show that this scheme can embed watermarks into important visual region, and consequently it can resist crop attack. Furthermore, the conception of this scheme is consistent with the MPEG4 and MPEG7 coding standard.