ExPACO: detection of an extended pattern under nonstationary correlated noise by patch covariance modeling

Abstract

In several areas of imaging, it is necessary to detect the weak signal of a known pattern superimposed over a background. Because of its temporal fluctuations, the background maybe difficult to suppress. Detection of the pattern then requires a statistical modeling of the background. Due to difficulties related to (i) the estimation of the spatial correlations of the background, and (ii) the application of an optimal detector that accounts for these correlations, it is common practice to neglect them.
In this work, spatial correlations at the scale of an image patch are locally estimated based on several background images. A fast algorithm for the computation of detection maps is derived. The proposed approach is evaluated on images obtained from a holographic microscope.