提出了一种基于堆栈滤波器和Hopfield神经网络的边界检测法，采用较小滤波窗口的堆栈滤波器优化估计的图象象素点之间的灰度梯度，再根据这些灰度梯度的优化估计值计算及确定Hopfield神经网络的权重矢量，Hopfield神经网络收剑时输出图象的边界。相对于基于堆栈滤波器边界检测法，该方法对堆栈滤波器的优化训练速度大大提高，所需内存大为减少，而相对于基于Hopfield神经网络的边界检测法，该方法又

A new kind of edge detection method based on stack filter and Hopfield neural network is proposed in this paper.First, the stack filter with smaller filtering window size is used to optimally estimate the gradient of gray scale for each pixels of the tested image. Then, weight vector of Hopfield neural network is determined by those optimal estimated values. Finally, converged Hopfield neural network outputs the image edges. Contrast to the stack filter based edge detection. the proposed method gainis higher speed, uses less memory for optimal training of stack filter. Also contrast to the Hopfield neural network based edge detection, the proposed method havs stronger abilities of reducing the effects of noise with different distributions, and can obtain much better edges.