Segmentation and detection of overlapped ellipses in binary images

This plugin proposes a complete method for recognizing and measuring highly overlapping ellipses in binary images. Starting from a binary image (the ellipses to be detected must be white on a black background), the pattern recognition process consists of the following successive steps:
i) identification of all the Region of Interest (RoI) - note that a RoI can be an isolated ellipse as well as a cluster of ellipses -,
ii) extraction of the edge of the whole RoI,
iii) splitting the edge into segments separated by connecting points,
iv) grouping the segments,
v) fitting an ellipse on each group of segment,
vi) candidate ellipses evaluation and bad candidates rejection.
Three parameters are used to tune the efficiency of the algorithm: the length of the concavity point detection segment and two ellipse rejection criteria (size and ellipticity) in the post-processing phase. They are to be adjusted according to the dimensions of the input images.
As an output, the algorithm returns an image of the detected ellipses and a text file containing their parameters in pixels (Center[x,y], semi-major axis, semi-minor axis and orientation).

This plugin requires the installation of the Matlab R2019a runtime. See the link below.

References:
[1] M. de Langlard, et al. “A multiscale method for shape recognition of overlapping elliptical particles”, International Conference on Pattern Recognition (2016).
[2] M. de Langlard, et al. “An efficiency improved recognition algorithm for highly overlapping ellipses: Application to dense bubbly flows”, Pattern Recognition Letters (2018).