基于神经网络的自适应图像分割

非均匀光照条件下的图像目标难以用阈值选取方法分割，本文使用神经网络建立自适应阈值曲面，以此作为图像分割的依据．自适应阈值曲面由图像中具有高Ｌａｐｌａｃｅ值的边缘拟合而成，拟合过程由神经网络实现．计算机摸拟证实了此设想的可行性以及相对阈值分割方法的优越性．

Adaptive Image Segmentation by Neural Network

It is difficult to separate an object from its background by means of thresholding in conditions of poor and nonuniform illumination. In this paper, we present a neuroncomputing approach for segmentation via adaptive thresholding. The threshold surface is determined by interpolating the image gray levels at points where the laplacian is high. The interpolation in the image plane is completed by a neural network. Through the experiments,it is shown that this method is better than the thresholding method.