Abstract: Multiview clustering frequently encounters missing views in practical applications, where existing methods struggle to balance the global consistency of low-rank graph structures with computational efficiency, and tend to underutilize consensus information under high missing rates due to neglecting geometric alignment among common samples across views. To address these issues, an incomplete multiview clustering method based on bilinear spectral penalty and structural alignment (SPIMC-SA) is proposed. Within an adaptive graph learning framework, a bilinear matrix factorization strategy replaces traditional nuclear norm constraints, transforming the low-rank subproblem from singular value decomposition into the optimization of smaller factor matrices to improve efficiency; a manifold regularization term then captures local geometric structure to compensate for the limited topology preservation of matrix factorization; and an explicit cross-view structural alignment mechanism uses common samples as anchors to enforce consistency of self-representation coefficients at the subspace level, calibrating distribution deviations between views. All variables are jointly solved under a unified ADMM framework. Experiments on multiple benchmark datasets show that SPIMC-SA achieves superior clustering performance with stronger robustness under high missing rates; for example, on BBCSport it achieves 93.25% and 78.62% accuracy at 10% and 50% missing rates respectively, providing an effective approach that balances solving efficiency and clustering robustness for incomplete multiview clustering under missing-data scenarios.