Comparison of image reconstruction methods for structured illumination microscopy

ABSTRACT Structured illumination micros copy (SIM) is a recent microscopy technique that enables one to go beyond the diffraction limit using patterned illumination . The high frequency information is encoded through aliasing into the observed image. By acquiring multiple images with different il lumination patterns aliased components can be separated and a high -resolution image reconstructed. Here we investigate image processing methods that perform the task of high -resolution image reconstruction, namely square -law detection , scaled subtraction, super -resolution SIM (SR -SIM), and Bayesian estimation. The o ptical sectioning and lateral resolution improvement abilities of these algorithms were tested under various noise level conditions on simulated data and on fluorescen ce microscopy images of a po llen grain test sample and of a cultured cell stained for the actin cytoskeleton . In order to compare the performance of the algorithms, the following objective criteria were evaluated: Signal to Noise Ratio (SNR), Signal to Background Ratio (SBR) , circula r average of the power spectral density and the S