Image filtering using anisotropic structure tensor for cell membrane enhancement in 3D microscopy

Image filtering methods are crucial in a large number of applications where the relevant information is either noisy or partially missing. More particularly in 3D microscopy, cell membranes appear as structures of co-dimension 2, and suffer from discontinuities in intensity and orientation. Anisotropic PDE-based approaches offer an elegant solution to these problems, by integrating spatial and orientation information in a flexible and robust manner. In this work we propose a new tensor-based anisotropic model specifically adapted to 2D planar structures able to perform noise reduction and contrast enhancement simultaneously. Evaluation on synthetic and real data as well as comparison with previous works show that the model is well adapted for cell membrane structures.