Super-Resolution 3D Reconstruction of Thick Biological Samples: A Computer Vision Perspective

In this paper we present a case-study about recent breakthroughs of three-dimensional (3D) super-resolution live-cell imaging through thick specimens (50 - 150um). This technology is enabling the deep understanding of cellular mechanism by obtaining very detailed 3D descriptions of cells. In particular, we discuss the image analysis problems related to the accurate localization of single molecules. This problem is hard because of the extreme noise conditions, the high and heterogeneous density of the cell molecules and the distortions induced by light-sample interactions on the imaging capabilities. For this reason, robust computational tools are required to obtain the localization of the photo-activated molecules and to enable the super-resolution accuracy. In such context, we show that a novel set of challenges exists and novel Computer Vision approaches are needed for delivering high-performing imaging systems for life science.