Interest point localization based on edge detection according to gestalt laws

The paper proposes a method for grouping fragments of contours of objects in the images of microscopic parasitological examinations, characterized by high transparency of analyzed objects. The method is based on a graphical representation of the edges in vector form, allowing to substantially reduce the required calculations. The method uses simple vector operations to determine stroke parameters describing the degree of curvature and closure of the examined contour and the direction where the remaining part of the contour should be located. Compared with other methods of detecting elliptic objects, the proposed method allows for the association of objects with rather irregular and distorted shapes occurring in parasitic images.