Thinning-like property of CNN with opposite-sign templates and its applications

A thinning-like (TL) property of cellular neural networks (CNN) with opposite-sign templates and a non-unity gain output function is discovered and proved. An image thinning algorithm is realized by a multi-layer CNN, which operates with the TL property with four local logic units (LLU). An image edge detection is achieved by another parallel CNN with a LLU. Some simulation results are given.