Computational removal ofbackground fluorescence for biological fluorescence microscopy

Background fluorescence is a frequently encountered problem in biological fluorescence microscopy. It often significantly lowers the image signal-to-noise ratio and poses substantial challenges to subsequent computational image analysis. Here we propose a general computational method for separating and removing background fluorescence from a single fluorescence microscopy image. The method is formulated as solving a constrained convex optimization problem and assumes that the background signal is low-rank and additive to the sparse foreground signal. Solution of the optimization problem is found using a forward-backward algorithm. Our method only requires a single image and can be used in a broad range of biological fluorescence applications. We first validate performance of our method using synthetic image data. We then demonstrate applications of the method to actual biological image data.