A DTCNN circuit proposal for pixel-level snakes

A VHDL description of a DTCNN circuit for pixel-level snakes is given. This is the first of successive steps in a top-down design flow towards a final physical implementation. The complexity of the application leads us to make use of a multilayer DTCNN with cyclic time variable cloning templates. In order to make a feasible physical implementation, the basic concepts of the CNN Universal Machine (CNNUM) have been adopted: distributed memory and programming templates. In addition, some other approaches like the use of 2Q multipliers are followed. The validity of the proposed structure is illustrated by simulations of a 9/spl times/9 network.

[1]  Josef A. Nossek,et al.  An analog implementation of discrete-time cellular neural networks , 1992, IEEE Trans. Neural Networks.

[2]  Xose Manuel Pardo,et al.  Pixel-level snakes , 2000, Proceedings 15th International Conference on Pattern Recognition. ICPR-2000.

[3]  Victor M. Brea,et al.  Discrete-time CNN for image segmentation by active contours , 1998, Pattern Recognit. Lett..

[4]  Tamás Roska,et al.  The CNN universal machine , 1993 .