Extraction method and denoising performance analysis of diffuse spots and silhouettes in STORM

In super-resolution microscopy based on STORM (Stochastic Optical Reconstruction Microscopy), due to the diffraction phenomenon, photons emitted by fluorescent molecules form into diffuse spots in a raw image. The regions without diffuse spots in the raw image are referred to as silhouettes. The diffuse spot regions are crucial for assessing the quality of a raw image and achieving super-resolution imaging as conventional holistic evaluation methods, which fail to distinguish between diffuse spots and silhouettes, cannot reflect the quality of the diffuse spot regions. A mask-based image segmentation algorithm is proposed here to extract and separate the diffuse spots and silhouette regions. Through simulation experiments, the denoising characteristics of three-dimensional block matching filtering (BM3D) and wide spectrum denoising (WSD) in different regions under various noise levels are compared and analyzed, and important criteria for selecting denoising methods for raw images are provided. Experiments show that better denoising is achieved with WSD than BM3D in the diffuse spot regions, while BM3D performs better in the silhouette regions in low noise level.