An interesting argument is presented in this paper, namely that the amount of data needed to describe a 3D object can be reduced with2D planes derived from the object with a computer. The paper does not, however, consider the latest technologies, such as parallelism or compression on a smaller machine, that could be used to compute the object.
Object recognition on a sequential machine, using invariant Euclidean signature curves from the 2D silhouettes of a 3D object, seems practical. It is quite possible that the experimental results of the signature curves will more closely resemble the silhouettes of the object when they are run on a parallel machine. Otherwise, the argument for the use of signature curves to describe the object seems practical, due to the type of machines (most likely sequential) the authors were using to make complex computations. These signature curves could serve as a starting point in the evolution of how to represent a 3D object in computer vision, as computer processor and compression technologies evolve.