摘要： 提出了一种基于纹理信息和遗传神经网络的图像分割方法。根据分形理论构造了图像的一组分形纹理特征参量,利用遗传神经网络作为图像像素的分类器,以此为基础对图像进行分割。分数维的计算采用改进的盒子计数法。试验表明,将该方法应用于生物医学图像,能够区分图像不同的纹理区域,达到满意的分割效果。

关键词: 纹理信息, 分数维, 遗传神经网络, 图像分割, 生物医学图像

Abstract: In this paper, a segmentation method is proposed by using genetic neural network and texture information. According to the theory of fraction, a group of fraction texture character parameters is obtained, and a genetic neural network is designed for classifying pixels of image, then the segmentation of image is made by using the neural network. The modified box-counting method is used in estimating fraction dimension. Results of experiment show that the proposed method can segment different texture area of biomedicine image in a satisfactory way.

中图分类号: 

• TP391.41