Abstract: A new method for image denoisng which combined quaternion wavelet transform (QWT) and adaptive bivariate model has been proposed. The adaptive bivariate model is used to describe inter-scale statistical correlation of the QWT decomposition coefficients. The Newton-Rapson is applied to estimate the marginal coefficient variance. Then the image denoising is done under the framework of Bayesian MAP estimation theory. The proposed algorithm outperforms in terms of denoising performance, peak signal to noise ratio, edge and texture information preservation.