A 3-D deconvolution based particle detection method for wide-field microscopy image

摘要

Tracking particles in fluorescent microscopy image is essential and challenging for many biological researches. The existing methods encounter difficulties and challenges for 3D dynamic process observation, especially for high density subcellular particles. In this paper we propose an automatic 3D detection method with wide-field fluorescence microscopy instead of confocal microscopy for high density particles. We employ deconvolution to improve the resolution, while the PSF of wide-field microscopy is measured by taking microscopic fluorescent beads as pointolites. For the complex situations of high density subcellular particles, we apply the matching scaled isotropic undecimated wavelet filtering and a global thresholding to find block candidates that contain at least one particle. Among the blocks we pick out the larger ones and perform an adaptive local thresholding based on 3D watershed. Experiments show that our method is of high accuracy and performs better than others.