An adaptive direction lifting multiwavelet algorithm is proposed, which combines the symmetry, orthogonality, short support set, high-order vanishing moment of multiwavelet with the directional characteristics of adaptive algorithm, therefore it improves the performance of image compression. The traditional multiwavelet transform only considers the vertical and horizontal direction transform, while adaptive lifting multiwavelet algorithm utilizes more local direction features of the image to look for the correlation between the pixels, so as to reduce the prediction errors of encoding. We use two D9/7 wavelets to construct a lifting multiwavelet, in combination with rate-distortion (RD) model to adaptively choose the lifting direction of the local region of the image, which makes a balance between reducing the prediction error and coding bit rate. Experimental results show that adaptive direction lifting multiwavelet has a better performance than the traditional lifting multiwavelet in image coding.