Focused on the coefficients′ interscale correlations of anti-aliasing contourlet transform (AACT),a hybrid research method which includes conditional probability,mutual information and mutual correlation and the generalized non-Gaussian bivariant distribution are proposed to model AACT coefficients. This model improves the solution method of shape parameter ξ. Experimental results of ξ and m1,1 (moment of order 1) show that the probability of ξ<1 is about 99.5% and the average value of error measurement of m1,1 is about 2.77%,which means this model is a good candidate for joint distribution of AACT coefficients.