Abstract: To address low-quality hexahedral meshes generated directly from industrial CT images, this paper proposes a boundary-constraint-free optimization method based on geometric element transformations. First, leveraging the characteristics of sequential CT images, an octree algorithm is employed to generate the initial hexahedral mesh. Then, to resolve boundary node misalignment and entanglement, a unitary geometric transformation method is introduced to adjust the positions of boundary nodes, ensuring their proper distribution. For distorted elements in the mesh, a regularization transformation is applied by incorporating the scaled Jacobian to improve the quality of hexahedral elements. Finally, experiments using CT images of actual workpieces demonstrate that the optimized hexahedral mesh achieves significant improvements: the minimum scaled Jacobian increases from −1.00 to 0.22, and the average scaled Jacobian rises from 0.89 to 0.93, making it suitable for practical engineering finite element analysis. These results validate the effectiveness of the proposed method.