Restoration of fluorescence images from two-photon microscopy using modified nonlinear anisotropic diffusion filter

Two-photon laser scanning fluorescence microscopy is becoming a powerful tool in study of neuron functional imaging in vivo for its inherent deeper penetration, less photo-damage. Now, with the two-photon fluorescence images of brain tissue, we can reconstruct three-dimensional neuronal morphologies easily. However, the images usually are obscured by a lot of noise, in particular in deep tissue with strong excitation laser power. Therefore, good image restoration technique that could remove the noise while preserve neuronal structure is crucial for the results of subsequent image segmentation and neuron reconstruction. Here, we propose a modified nonlinear anisotropic diffusion filter which incorporates both gradient and gray-level variance of raw data, to remove the noise, rather than merely considers gradient as the classical Perona-Malik nonlinear anisotropic diffusion model. Experimental results have shown that the proposed scheme can remove noisy speckles effectively while maintain the shape of neuronal morphologies in two-photon fluorescence images without conflict.