摘要:
为了简单有效地提取图像重要特征信息, 从而更好地提高检索图像的精度, 提出了一种基于脉冲耦合神经网络(Pulse coupled neural networks, PCNN)的图像归一化转动惯量(Normalized moment of inertia, NMI)特征提取及检索算法. 首先利用改进简化PCNN模型相似神经元同步时空特性及指数衰降机制将图像分解为具有相关性的二值系列图像, 然后提取反映原始图像目标形状、结构分布二值系列图像的一维NMI特征矢量信号, 并将其应用在图像检索中; 同时, 考虑到二值系列图像间的相关性及不同图像间NMI序列值的差异性, 引入了马氏距离结合Pearson积矩相关法的 综合相似性度量方法. 实验结果表明, 所提算法对图像特征矢量序列具有良好抗几何畸变不变特性及对图像表述的唯一性,且具有较好的图像检索效果.
Abstract:
In order to simply and effectively extract the information of important features in the image so as to improve the accuracy of the image retrieval, a novel algorithm of image normalized moment of inertia (NMI) feature extraction and retrieval based on pulse coupled neural networks (PCNN) is put forward. Firstly, the image is segmented into a series of binary correlation images using synchronous spatial-temporal characteristics of similar neurons and exponential attenuation mechanism of improved and simplified PCNN, and then a one-dimensional NMI feature vector signal of the binary series images, which can reflect the target shape and structure of the original image, is extracted, and applied to the image retrieval. Meanwhile, considering the correlation between binary series images and NMI sequence values' differences between different images, the method of compounded similarity measurement of the combination of Mahalanobis distance and Pearson product-moment correlation is introduced. Experimental results show that the proposed algorithm has good performance of anti-geometric distortions and the uniqueness for different images' expression to the vector sequence of image features, and has better image retrieval results.
下载:
