To further enhance the performance of existing watermarking schemes based on the support vector machine(SVM),an image watermarking scheme in contourlet domain based on support vector regression(SVR) is proposed in this paper.Firstly,the host image is decomposed by the contourlet transform.Then,the non-linear local correlation model of the image is established using the support vector machine.The watermark is adaptively embedded according to the prediction result of the established model.Experimental results show that the proposed scheme is not only invisible and robust against the common image signal processing,such as noise adding,JPEG compression,sharpening,smoothing filtering and contrast enhancement,but also robust against the geometric attacks,such as rotation,shearing and distortion.Especially,its performance is significantly superior to the image watermarking schemes in spatial domain or wavelet domain based on the support vector machine.