A texture reconstruction method based on Markov random fields optimization was proposed to reconstruct the complex free-form shapes. The method used a three-dimensional digital device developed by laboratory to capture the range data and the texture images, and registered the local capture range data into global coordination system to establish the surface of the object. To build a realistic textured model, the captured texture images were mapped to the reconstructed surface by coordinate transformation, and a texture fusion processing was followed. The proposed method does not impose any constraints on the topology of the surface, and a realistic textured model of the free-form surface can be achieved. The data collection and realistic three-dimensional reconstruction of material object were experimented by the proposed method, the results show the reliability and effectiveness of the proposed method.