Adaptive watermarking of color images in the DCT domain

This paper presents a new approach of embedding watermarks adaptively in DC components of color images with respect to analysis of the image luminance in the YIQ model or the blue channel in the RGB model. The embedding strength is determined by the classification of DCT blocks, which is simply implemented in the DCT domain via just analyzing the values of two AC coefficients rather than using method such as Sobel, Canny, or Prewitt. Furthermore, a new combination algorithm of both watermark extraction and blind detection is designed, where the watermark is extracted directly in the spatial domain without knowledge of the original image. Experimental results demonstrated that the embedded watermark is robust to the attacks including JPEG-loss compression, JPEG2000, filtering, scaling, small cropping, small rotation-crop and self-similarities attacks.