Detection and Characterization of Isolated and Overlapping Spots

摘要

A systematic approach is presented to detect and characterize isolated and overlapping spots. In defining spots as image details without inner structure, the analysis reduces to the determination of their location, intensity, size, and possibly orientation. The detection phase consists of matching a spot model to all image positions. The characterization phase consists of searching the best matching model for each spot. Both matching procedures are based on weighted image matching, where the weights decrease at greater distances to reduce the influence of other, proximate spots. Experiments on signal-to-noise ratio, parameter accuracy, and the smallest distance which still allows separate detection of two spots, show that the weight function should be large for isolated spots but small for overlapping spots. The method is put into practice by detecting and characterizing stars on a nebular background and clustered yeast colonies in a Petri dish. The weighted spot matching method allows accurate spot characterization while the spot definition is transparent and flexible. In contrast to other methods, our approach is consistent among the detection and the characterization phase. It will separate overlapping spots even if they are showing one intensity maximum, when the appropriate size is known.