Robust subspace clustering image segmentation algorithm based on noise suppression

Separating meaningful visual structures from images is one of the most fundamental tasks addressed by image analysis algorithms. Research on existing subspace clustering segmentation methods reveals that most of them can only handle noise-free or low-noise scenarios. The proposed method aims to address the robustness issues of subspace clustering segmentation algorithms in image segmentation tasks under high noise conditions. Image segmentation, as the subspace clustering of image feature vectors, analyzes the impact of noise on subspace clustering segmentation algorithms. Firstly, the concept of non-local means filtering is applied to enhance the precision of image segmentation and to improve its resistance against high-noise segmentation models. Secondly, a novel centroid update robust superpixel segmentation algorithm is employed along with adaptive image denoising theories to reduce the noise influence on superpixel feature extraction, thereby producing more accurate feature vectors. Multiple features are extracted to establish a feature data matrix. This paper integrates these three denoising concepts and multiple superpixel feature extractions for the comprehensive application in subspace clustering segmentation algorithms. This approach aims to enhance the robustness of segmentation algorithms under noise and produce correct segmented images.