Abstract: This study proposes a new algorithm to enhance the accuracy of measuring internal deformation, identify and quantify damage in rocks using digital volume correlation (DVC). This algorithm integrates the multimesh refinement technology into a mechanically regularized global DVC algorithm that incorporates nodal force equilibrium. The accuracy of different DVC algorithms was assessed through scanning experiments on particles with different sizes and digital virtual fracture models. The internal deformation and damage patterns of red sandstone were quantitatively analyzed using in-situ CT scanning during uniaxial compression experiments. The results indicate that the introduction of the mechanical regularization term reduced the measurement uncertainty of the global DVC method by 1-2 orders of magnitude. For the red sandstone specimens, the new algorithm was able to identify sub-voxel microcracks with an aperture of 0.15 voxels, while reducing the computational cost by 85.15%. The damage quantification analysis revealed that newborn cracks with a maximum aperture of 0.32 voxels occurred in the specimen when the axial stress reached 50.58% of its peak value, and these microcracks evolved into voxel-scale fractures when the stress increased to 84.27% of the peak value. The new algorithm not only improves the computational efficiency, but also demonstrates a strong capability in identifying sub-voxel scale newborn cracks, providing a novel method for quantifying internal deformation and damage in rocks.