Locating neuron boundaries in electron micrograph images using 'Primal Sketch' primitives.

This paper presents an approach to this problem that uses "Primal Sketch"-like computations from David Marr's vision framework. The computations are line averages computed over the image and stored symbolically in a table. These symbolic lines can be manipulated using geometric constraints to locate the boundaries of neurons. A graph search algorithm is used in this preliminary study to demonstrate the potential of the Primal Sketch, and neuron boundaries located that would be difficult or impossible by iso-density contour following. This is a first step to full reconstruction of the 3D tissue structure from a series of sections.