Aluminum foam is a type of cellular materials generally characterized by high porosity; it contains a high fraction of pores and enclosed voids. The accurate thermal conductivity of aluminum foam plays a very significant role in using and designing foam materials. In this paper, a novel method of reconstructing closed-cell aluminum foam by using MATLAB image processing and CT scanning is put forward. First of all, the maximum entropy algorithm is employed to obtain the binarization image. Secondly, median filtering algorithm is used to reduce the "salt and pepper" noise after binarization. Thirdly, the external contour and internal pore boundary is extracted by the "edge" function in MATLAB. Then, geometrical model of the foam plate is reconstructed. Finally, a two-step mesh algorithm is used to mesh the reconstructing geometrical model. Comparing the illustrated results with experiment data, the Parallel-Serial Model can better calculate the thermal conductivity of closed-cell aluminum foam at different porosity. Unlike Michailidis's 3D reconstruction method and Jeon's 3D reconstruction model, the novel method has a smaller difference in structure size between the geometric model and the real model, and higher efficient. The calculation results agree well with that by the Parallel-Serial Model based on the reconstruction geometry model, and the reconstructed geometry can be used to discuss the influence of cell size on the thermal conductivity of aluminum foam, which proves the feasibility of the geometrical method for research on the thermal conductivity of closed-cell aluminum. Thermal conductivity of closed-cell foam depends on both of the porosity and the cell size of pores. (c) 2014 Elsevier Ltd. All rights reserved.