Microstructure characterization and reconstruction (MCR) can be used to establish the processing–structure-–property (PSP) relationship, which has become the focus of computational materials science and material design. In this paper, threshold segmentation is used to identify the metallographic photos of heterogeneous materials as textures composed of different phases, combined with multi-resolution Gaussian pyramids for reconstruction. The reconstruction results after particle number supplement are iteratively optimized with percentage of each phase as target parameter. Thereby, this paper proposes an innovative MCR texture synthesis algorithm with controllable area fraction, which is used to reconstruct the microstructure of heterogeneous AlSi–PHB seal coating. Two-point correlation functions of PHB phase and pore are utilized to confirm the effectiveness of the proposed algorithm. On this basis, two sets of metallographic photos are used to reconstruct 20 sets of models, and the maximum error of the two-point correlation function of the reconstructed results is 8.8×10^–5 . This paper provides a solution for the accurate reconstruction of heterogeneous materials.