A versatile major axis voted method for efficient ellipse detection

The recent developments in the field of particle science and technology show that the shape of the particle other than size and surface chemistry are important. Most commonly, digital images captured either by microscopy or other means are analyzed by image processing methods to completely characterize the shape of the objects. Among many possible shapes, ellipse like objects are ubiquitous and encountered across several disciplines - for example - face recognition in computer vision, classification of food grains and leaves, study of particle shape effects in colloid science, etc. Detection of particle position and their orientation is therefore of great interest to quantitatively understand the phenomena under investigation. In this article we propose a major axis voted method for the detection of ellipse-like objects in 2D images which is based on the assumption that each pair of boundary pixels are on the endpoints of the minor axis of the ellipse. We show that the method thus developed is capable of detecting occluded ellipses as well as deformed ellipses.