利用PCNN实现商标图像检索新方法

 为了解决商标存在尺度缩放、角度旋转和区域局部变化后的难以检索的问题,提出了一种基于脉冲耦合神经网络模型(PCNN,Pulsed Coupled Neural Network)的商标检索新方法.首先通过PCNN图像分割和直方图均衡化技术相结合提出了一种有效的图像预处理算法,以减小颜色对于商标灰度分布差异产生的影响.然后在PCNN模型中提出了边缘时间序列概念,并用于提取商标图像的形状特征,最终实现了商标图像的快速有效检索.在所建立的商标库中进行了计算机仿真,仿真结果表明该方法可有效地检索出待检索商标对应的商标图像,可很好地适应商标颜色变化、角度旋转和局部形状变化,体现了较好的检索性能.

A new method of trademark image retrieval using PCNN

In order to solve the scaling,rotation and local changes in the trademark image retrieval.A method of trademark retrieval based on pulse coupled neural network(PCNN) model is proposed in this paper.First an effective image preprocessing algorithm,which reduces the influence of trademark gray distribution difference by the trademark color,is presented by combining PCNN image segmentation and the histogram equalization technology.Then the edge time sequence concept,which is used to describe the distribution characteristics of trademark images regional edge in space,is put forward based on PCNN model,and finally achieve the trademark images quickly effective retrieval.We proceed to the computer simulation based on the trademark repository,simulation results show that this method can be entirely accurate to retrieve the corresponding trademark image,so the method has great resistance to color trademark transformation,rotation,local shape changes and good retrieval performance.