Fast and robust pattern detection: Application to spherical bead location in holographic microscopy

In several areas of imaging, it is necessary to detect the weak signal of a known pattern superimposed to similar or more contrasted patterns. Detecting the pattern of interest then requires robust detection techniques. In the case of spatially extended patterns, the computational complexity of systematically evaluating a robust loss function at all possible locations becomes prohibitive. In this work, we propose a fast detection algorithm applicable to various robust loss functions. The proposed approach is evaluated on simulated and experimental in-line holograms.