A three-dimension multi-scale fusion reconstruction method for porous media based on pattern-matching

Accurate three-dimensional modeling has an important influence on the evaluation and prediction of the physical properties of porous media. In view of the inherent natural structures, pores are multiscale in many porous media. As a result of the tradeoff between the field of view (FOV) and resolution, it is difficult to obtain high-resolution and large-scale images using single-resolution imaging equipment. Generally, low-resolution images with a large FOV lack detailed features, while the FOV of high-resolution images is too small to contain the information of the entire microstructure. Therefore, fusing images from different resolutions and scales to reconstruct models is an promising solution for addressing these issues. In this study, an improved method for reconstructing a high-resolution and large-scale model based on three-dimensional pattern matching is proposed. The algorithm can fuse two models from different resolutions and scales, in which the structures of large-scale pores in the low-resolution model are used as background and a high-resolution model is applied to build pattern sets to reconstruct fine-scale pores based on the background. The final result is a model containing large-scale and fine-scale pores. The accuracy and stability of the proposed method were tested by using two samples. From subjective and objective perspectives, through the comparisons of visual effects and statistical properties between generating models and target models, we found that the method can reconstruct accurate models.