Fig. 1. SMLM imaging process and SMLM image characteristics. (a) Conventional fluorescence image; (b) SMLM data acquisition; (c) point cloud image reconstruction; (d) point cloud image rendering; (e) enlarged view of selected region in rendered image; (f) SMLM spatial resolution

Fig. 2. Flowchart of image segmentation by combination of multi-step DBSCAN and hierarchical clustering algorithm. (a) Input of original data; (b) the first DBSCAN segmentation; (c) cluster area analysis and threshold setting; (d) super-threshold area clusters extraction; (e) secondary DBSCAN segmentation and hierarchical clustering; (f) output of segmentation result

Fig. 3. Influence of parameter selection on clustering performance. (a) Ground truth of D31 dataset; (b) influence of selecting too small parameter on improved clustering algorithm; (c)-(f) influence of different parameters on performance of traditional DBSCAN algorithm

Fig. 4. Clustering segmentation of simulated datasets. (a) Ground truth of D31 dataset; (b) segmentation of D31 dataset by traditional DBSCAN method; (c) segmentation of D31 dataset by improved clustering method; (d) ground truth of S2 dataset; (e) segmentation of S2 dataset by traditional DBSCAN algorithm; (f) segmentation of S2 dataset by improved clustering algorithm

Fig. 5. Clustering segmentation for SMLM image of membrane proteins. (a) Segmentation effect of uniform distributed membrane protein SMLM data; (b) Ripley's K function of uniform distributed membrane protein SMLM data; (c) segmentation effect of random distributed membrane protein SMLM data; (d) Ripley's K function of random distributed membrane protein SMLM data; (e) segmentation effect of non-uniform distributed membrane protein SMLM data; (f) Ripley's K function of non-uniform distributed membrane protein SMLM data