Robust Color Image Watermarking Approach Based on Shape-specific Points

In this paper, we propose a watermarking scheme with high perceived quality for color images by using the Euclidean distance of radius-weighted mean (RWM) and centroid. Based on spatial domain, the proposed watermarking scheme can effectively embed a secret message in host images. Simulations demonstrate that the marked images generated by the proposed watermarking scheme are tolerant of versatile attacks such as color reduction, compression, cropping, lens-blurring, noise addition, pixel-truncation, and so on. Above all, the extracted watermarks are recognized when the marked images suffered from attacks such as JPEG2000 compression with a high CR attack and large-scale noise-addition attacks.