摘要：In this paper, a segmentation algorithm foratomic force microscopy images has been developed, usingwave region growing around local maxima as a result ofjoining adjacent pixels to the regions, which are selected indecreasing order of values. The essence of the algorithmconsists in using the brightness threshold, which is graduallychanging from maximum to minimum, in order to selectgrowing points or to join existing areas. The iterativelyexpandable boundaries, and the choice of initial growing pointsand points that are attached to areas with a focus on thresholdvalues with a gradual decrease from maximum to minimum,are considered features of the developed segmentationalgorithm. These features made it possible to eliminate errorsof marker watershed, region growing algorithms andwatershed using the classical Vincent-Sollie algorithm, whichare usually used in segmentation of AFM-images. Thedeveloped algorithm is compared with the following standardalgorithms: classic watershed algorithm, marker watershed,region growing. The comparison is made on test and originalAFM-images. The algorithms are implemented in Matlab andC ++. To quantify segmentation errors, a set of binary masks isused. The experiments showed that the developed algorithmselects the region boundaries without errors and with highersegmentation speed in comparison with the algorithms forregion growing and Vincent – Sollie watershed.
关键词：image segmentation; atomic force microscopy;
region growing; Vincent–Sollie watershed; local maximum;
AFM-images; marker watershed
