The pore structure of the filter cake plays a crucial role in the filtration efficiency. However, the existing three-dimensional characterization techniques are faced with high testing costs, which limits the in-depth understanding of the dynamic dewatering mechanism. This paper conducts a study on the three-dimensional reconstruction technology of the pore morphology of industrial coal sludge filter cakes by using ImageJ software in combination with multi-point sampling of the filter cake and scanning electron microscope examination (SEM). In addition, by cutting the three-dimensional reconstructed filter cake slices and using the Dijkstra algorithm for path analysis, the porosity gradient (20.0 - 43.6%), ineffective pores (2.19 - 9.49%), and tortuosity (1.64 - 2.05%) at the micrometer scale resolution were quantified. The results show that coarse particle settling increases the bottom porosity, while fine particles near the feed inlet increase the ineffective pores, resulting in a 30.6% reduction in effective permeability. This method establishes a spatial connection between the heterogeneous structure of the filter cake, pore connectivity, and dewatering efficiency. Compared with the traditional two-dimensional theoretical model, this method underestimates the total porosity by 12 - 18%, emphasizing the necessity and effectiveness of three-dimensional analysis.