A Local Watermarking Scheme in the Ridgelet Domain Combining Image Content and JND Model

In this paper, a local watermarking scheme in the ridgelet domain combining image content and JND model is presented. Since the ridgelet transform (RT) can efficiently represent image with linear singularities and has directional sensitivity, the image is partitioned into small pieces. And these small pieces are classified into different characteristic categories (S1 with weak texture, S2 with strong texture). To improve the robustness of processing like noise or JPEG compressing attack, the lower and middle frequencies of RT subband are selected and the watermarking strength is adaptively determined by the JND values in RT domain. So it will ensure invisibilities of the watermarks. And eventually the watermarks are embedded in the most important energetic directions of the pieces with strong texture which are less sensitively to human¿s vision. Experimental results show that the proposed watermarked scheme is robust to noise, cut, JPEG compressing and other intensive attack.