Incomplete multiview clustering (IMVC) has attracted extensive attention in the field of machine learning due to its excellent performance in handling incomplete multiview data. However, existing IMVC methods have certain limitions:1) the completion methods are not flexible enough for cases where view information is arbitrarily missing; 2) the existing methods cannot fully exploit the information contained in the clustering indicator matrix, which is crucial to the clustering results. To solve these problems, we propose a novel method, i.e., tensor-based global block-diagonal structure radiation for incomplete multiview clustering (TGBSR). First, we utilize the information from known samples to construct similarity graphs and integrate these matrices into a third-order tensor. Second, to explore the consensus information among the different similarity graphs, we introduce the spectral embedding technique to obtain the clustering indicator matrix, i.e., U. Third, and more importantly, to fully explore the information contained in the clustering indicator matrix, we introduce a novel matrix, i.e., the UUT, which is low-rank and naturally has an excellent block diagonal structure, and then incorporate it as a new view into the aforementioned tensor, which is constrained by the tensor nuclear norm (TNN), to mine the high-order correlations among all the different views. Finally, we combine all these steps into a unified framework and develop an effective iterative optimization procedure to solve it. The experimental results on several well-known datasets demonstrate the effectiveness of our method. The code is publicly available at https://github.com/hhipp-999/TGBSR.