Abstract: An image denoising method using complex Gaussian scale mixtures(CGSMs) model in complex Curvelet transform(CCT) domain is presented.The reason of "scratches" and "embedded stain" in the Curvelet transform reconstruction image is pointed out that aliasings exist in Curvelet transform domain.So,a new non-aliasing Curvelet transform,namely CCT,is proposed by using complex wavelet transform and improved Radon transform to replace real wavelet transform and old Radon transform in the original Curvelet transform respectively.The Gaussian scale mixtures(GSMs) model is extended into complex wavelet domain and becomes a CGSM model which can capture both magnitude and phase information of the complex wavelet coefficients availably.In the CCT domain,the noisy coefficients in the CGSM model can be estimated effectively by using Bayesian least squares(BLS) estimator,so noise reduction can be achieved.Experimental results show that,whether judged by PSNR index or in visual effect,the proposed scheme outperforms the traditional Curvelet transform denoising,Curvelet domain HMT denoising and wavelet domain BLS-GSM denoising.Plus,it has a good ability to preserve image details and edges while suppressing noise effectively.