基于神经网络与遗传算法的纹理图像分割

Segmentation of Texture Image Using Neural Network and Genetic Algorithms

运用两维自动回归模型、分形维数、均值和方差从每一小区域的数据中抽取纹理特征 ,把纹理特征作为自组织特征映射神经网络的输入层进行训练确定最优的纹理区域分割数 ,最后运用遗传算法优化图像分割。

Texture features are extracted from the data in each small region by using two dimensional autoregressive model, fractal dimension, mean and variance of the pixel data. These texture features are considered as the input layer of self organizing feature map, then the optimum number of segmentation areas are confirmed by training the neural network. Finally using GAs optimizes image segmentation. The results show the method can implement texture image segmentation effectively.