A novel SVD and LS-SVM combination algorithm for blind watermarking

This paper proposes a novel algorithm for blind watermarking by applying singular value decomposition and least squares support vector machine into watermark embedding and detection. In coding process, singular value decomposition is performed on coefficient blocks to obtain singular values after host image is transformed into integer wavelet transform domain. Subsequently, watermark image is embedded into transformed image by adaptively modulating the sample feature vectors constructed by singular values. In decoding process, the trained least square support vector machine is employed to extract the watermark image blindly by classifying samples derived from watermarked image. Experimental results show that the proposed scheme is not only robust against common noise-like attacks, such as noise, filter, crop, sharpen and JPEG compression, but also robust against geometrical distortions.