A neural network for learning Hough transform for conoidal structures

A 2-layered neural network, namely, Hough transform network, is designed to learn parametric forms of conoidal shapes (e.g., lines/circles/ellipses) from images and higher dimensional input. It provides an efficient representation of visual information embedded in the connection weights and parameters of the processing elements. It not only reduces the large space requirements of classical Hough transform, but also represents parameters with a higher precision