Robust subspace clustering via joint reweighted Frobenius norm and Lq norm minimization

Subspace clustering minimization is a common problem in machine learning and image segmentation. Nevertheless, the classical nuclear norm exacerbates the impact of significant singular values. This paper presents a novel approach for clustering incomplete noise observation data by introducing a joint reweighted Frobenius norm and norm model. Subsequently, we present a new algorithm for resolving the model. In addition, we incorporate a weight matrix to flexibly modify the influence of significant singular values. The two datasets ultimately confirm that the Joint Inexact Splitting Method (JISM) exhibits superior clustering segmentation efficacy and ensures algorithm convergence. Furthermore, JISM has good performance in restoring block occlusion of any size.