用神经网络算法进行纹理分割

NEURAL NETWORK ALGORITHM FOR TEXTURE SEGMENTATION

本文使用神经网络多层感知机模型，利用误差逆传播学习算法，结合纹理分析的基本手段，通过对计算机模拟神经网络进行训练，使它掌握了多种纹理的纹理特征，有效地实现了对包含多种自然景物纹理及多种人工图案纹理的图象的分割。

In this paper, we employ the multilayered perceptron neural network of which the error BP algorithm is the Learning principle to segment an image with five kinds of natural textures and an image with six types of artificial designs. The texture analysis which is employed by our experi- ment to support the neural net work segmentation is based on SOLD method, in which one of the SOLD derived features-correlation is used to describe texture..The segmentation results are quite satis- factory.