Current Issue Cover
基于神经网络的纹理图象生成

吴小培1, 周荷琴1, 冯焕清1(中国科学技术大学电子科学与技术系,合肥 230026)

摘 要
提出了一种利用多层前馈神经网络生成纹理图象的新方法.利用该方法可方便地生成图案丰富的纹理图象集,并且该纹理图象集中的任何一幅图象均唯一地对应一组神经网络的权值和阈值,因此不仅便于图象保存,还大大地节省了图象存储空间.
关键词
The Texture Image Generation Based on Neural Network

()

Abstract
In this paper, a new method to generate the texture image by use of Multi-layer feed-forward neural network is presented. This method can be used conveniently to generate different kinds of texture image sets. Each image in texture image sets corresponds to a group of definite weights and thresholds of neural network. This characteristic of the method is very useful for image' s saving and then a lot of storage space can be saved.
Keywords

订阅号|日报