<p>Multi-view subspace clustering (MVSC) arms to uncovers shared low-dimensional structures across views. However, real-world data often contain noise and missing data, which hinder the effective use of complementary information and degrade clustering performance. To address these challenges, we propose DMFGF-MVSC, a novel MVSC method that integrates double-constrained matrix factorization with graph filter. The graph filter enhances latent representations by smoothing multi-view data and suppressing redundant features. Meanwhile, the double-constrained matrix factorization decomposes each view’s self-representation matrix into basis matrices and a shared encoding matrix, promoting consistency and complementarity across views. Furthermore, to handle incomplete data, we introduce binary mask matrices that enables DMFGF-MVSC to remain effective in missing-data scenarios. Extensive experiments on seven benchmark datasets show that DMFGF-MVSC outperforms nine state-of-the-art complete multi-view clustering methods. Additionally, on three simulated datasets with missing views, our method surpasses three representative incomplete multi-view clustering approaches. These results highlight the effectiveness and robustness of DMFGF-MVSC for both complete and incomplete multi-view subspace clustering tasks.</p>