Both the Directionlet and Contourlet transform are efficient transform-based representation for image processing. However, the design of the multidirectional filters in Contourlet always turn out high complexity; and not as the wavelet transform, the Directionlet don’t retain the resolution-scalability, which leads to weak expressions of the smooth regions in images. To overcome these drawbacks, the Laplacian Pyramid in the Contourlet and the Directionlet is combined and Contour based Directionlet transform is constructed. Numerical experiments are performed to assess the applicability of the proposed method. The obtained results show that the proposed method could well express both smooth regions and details in image, and obtain higher metrics compared with Wavelet, Contourlet and Directionlet transforms.