Abstract: X-ray computed tomography (X-CT), a non-destructive and high-resolution three-dimensional imaging method, is widely used in clinical medicine and industrial inspection. In the field of geoscience, X-CT has been used to digitally model oil and gas cores and has potential for analyzing pore structures and seepage characteristics of digital cores. However, research on internal imaging of ore-bearing rocks using this method remains relatively limited. In this study, banded niobium-rare earth-fluorite ore with a complex composition and two mineralogically similar rock cores were selected as research objects. X-CT was used for scanning imaging, and a series of two-dimensional grayscale images were reconstructed using the cone-beam projection reconstruction (Feldkamp–Davis–Kress) algorithm. These images were segmented based on the density differences of the samples and internal information extraction models for ore-bearing rocks were established. The volumes of the samples and each mineral component were analyzed statistically. The following conclusions were drawn: (1) X-CT demonstrated high adaptability to the geometric morphologies of the rock samples. Taking these rock samples as a representative case, the industrial CT system employed achieved a maximum spatial resolution of 45 μm for sample scanning. (2) The grayscale of the segmented images was represented by 16-bit grayscale values. A comparison of the grayscale ranges of the samples showed that rocks with complex mineral compositions and similar densities exhibited similar grayscale values for internal minerals, whereas for samples with relatively simple mineral compositions and significant density differences, the grayscale values showed a wider distribution range. (3) The analysis results of the two mineralogically similar rock cores were compared, revealing that the same mineral exhibited different grayscale ranges for different cores scanned under different parameters. In contrast, for different minerals in the same core, the greater the density difference, the more significant the difference in the grayscale images and the easier it was to extract.