An edge detection method based on feed forward neural network and Hopfield neuralnetwork are proposed

the feed forward neural network is used to enhance and encode the information in the region of tested pixel

and its result is the input of Hopfield neural network which does iterative working till the network converges. The new neural network saves the computational cost and can generate edge map clearer than traditional Hopfield neural network. The whole system work sun der unsupervivsed leaning

so it is easy to train the neural networks.