Skip to Main Content
In this paper we present a robust image watermarking method against geometric distortions and several attacks of signal processing. Initially, the original image is normalized and the feature points are then extracted. Furthermore, the locally most stable feature points are obtained and used to generate several non-overlapped circular regions. Texture-based image block classification in Discrete Cosine Transform (DCT) domain is also used to get watermark imperceptibility. The watermark is embedded and extracted into classified regions using the DCT domain. In the detector stage, the watermark can be obtained without the original image. The experimental results show the watermark robustness against geometric distortion such as rotation, scaling and various other Affine Transformation, and common signal processing attacks, such as JPEG compression and noise contamination.