基于嗅觉神经网络仿生模型的纹理图像生成算法(英文)

Algorithm for Texture Image Generation Based on a Bionic Model of Olfactory Neural Networks

提出了一种基于嗅觉系统生成纹理图像的仿生模型。该模型结构模拟嗅觉神经网络的结构,利用Logsitic函数的混沌特性调整每次迭代过程中的模型参数,使用简单的周期函数作为模型节点的激活函数实现纹理的重复,并引入随机噪声来模拟脑在进行信息处理时的背景噪声。实验结果表明,该模型可以生成丰富而多变的纹理图像,引入的随机噪声也起到了积极的作用,可以明显地丰富纹理图像的变化。此外,模型生成纹理图像的效率也高于传统的BP神经网络模型。

This paper presents a novel bionic model based on olfactory systems to generate texture image. The model simulates one of the olfactory neural networks. The chaotic characters of Logistic function are used to adjust the parameters of model during iteration. A simple periodic function is used as the activation function of node in the model to generate periodic texture. And a random noise is introduced to simulate the background noise of brain when processing information. The experimental results show that the model can generate plentiful and multivariant texture images. The introduced random noise plays an important role and enriches the variety of texture images obviously. In addition, the model efficiency to generate texture image outperforms the conventional back propagation neural network model.