One of the earlier approaches to deformable template analysis
[Yuille91] was aimed to find facial features for purposes of recognition. The approach considered an eye to be comprised of an iris which sits within the sclera and which can be modelled as a combination of a circle that lies within a parabola. Clearly, the circle and a version of the parabola can be extracted by using Hough transform techniques, but this cannot be achieved in combination. When we combine the two shapes and allow them to change in size and orientation, whilst retaining their spatial relationship (that the iris or circle should reside within the sclera or parabola), then we have a deformable template. ...