Combination of Fast, Robust TV-based Deconvolution with Variational Segmentation for Microscopic Data

• Applicant: Professor Dr. Jürgen Hesser
• Subject Area: Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
• Term: from 2005 to 2011
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 14024299

Quantitative measurements in biology using microscopy allow understanding of cellfunctions. The necessary basis is the reliability and sensitivity of such a measurement in the case of low photon counts since a low photon count is required in order to avoid photo bleaching or damages to the cell (phototoxity). In this project, we will implement such a robust measurement system for fluorescent microscopy. It will for the first time combine optimal methods for PSF estimation, segmentation, and deconvolution. We will develop a method for PSF estimation using Zernike polynomials that are adapted to both, the degrees of freedom of physically feasible PSFs, and the physical noise model of the system. We extend this method by using a combined segmentation-blind deconvolution approach, restricting the PSF estimation to areas where we segment objects of known structure properties. This combination should not only lead to better results, we will also overcome known limitations of blind deconvolution, i.e. the insensitivity to the PSF kernel. To make the system applicable in practice, new approaches to speed up computations are implemented, e.g. using tube methods, adaptive restoration and ordered subsets. We expect a significant improvement in accuracy and reliability compared to the state of the art.

• DFG Programme: Research Grants
• Participating Person: Professor Dr. Reinhard Männer