脉冲耦合神经网络纹理检索系统研究  

Texture Retrieval System by Using Pulse-Coupled Neural Network

作　　者：杨丽[1] 

机构地区：[1]新疆职业大学招生与就业指导办公室,新疆乌鲁木齐830013

出　　处：《新疆职业大学学报》2011年第3期76-80,共5页Journal of Xinjiang Vocational University

摘　　要：脉冲耦合神经网络用于图像检索时,具有平移、旋转、尺度、扭曲等不变性,同时对噪声还具有很好的鲁棒性.本文提出了一种改进的脉冲耦合神经网络图像纹理检索算法,该算法将脉冲耦合神经网络和离散余弦变换相结合.首先将脉冲耦合神经网络输出的二值图像序列进行离散余弦变换,然后选择不同的特征提取方法提取纹理图像的关键特征,最后利用欧氏距离分类器对图像进行识别.基于Brodatz纹理图像库的实验结果证明了该算法的有效性。Pulse-Coupled Neural Network, which is used for image retrieval, takes on invariance of translation, rotation, dimension, distortion, etc. At the same time, it has robustness to noise. In this paper, an improved image texture retrieval algorithm by using Pulse-Coupled Neural Network is presented, it combines Pulse coupled neural network with discrete cosine transform. First, it can do discrete cosine transform on the binary image sequence out from Pulse-Coupled Neural Network. Then it can get texture image feature by different feature extraction methods. Finally, it can recognize the images by Euclidean Distance classifier. The experimental results based on Brodatz＇s texture image database shows its effectiveness.

关 键 词：纹理检索 特征提取 脉冲耦合神经网络 交叉皮层模型 离散余弦变换 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]