基于小波变换的PCNN多传感器图像融合  

Fusion of multi-sensor images based on wavelet transform and PCNN

作　　者：薛寺中[1] 周爱平[2] 梁久祯[2] 

机构地区：[1]江南大学数字媒体学院,江苏无锡214122 [2]江南大学信息工程学院,江苏无锡214122

出　　处：《计算机工程与应用》2011年第35期210-213,217,共5页Computer Engineering and Applications

基　　金：江苏省自然科学基金(No.BK2008098);教育部留学人员回国启动基金项目

摘　　要：利用PCNN(Pulse Coupled Neural Network)在图像处理中的独特优势,提出了一种基于小波变换的PCNN多传感器图像融合方法。对源图像进行小波分解,得到不同尺度下的子带图像;在小波域中利用PCNN的同步脉冲激发特性,制定基于PCNN的融合规则;使用不同尺度下的小波系数的SF(Spatial Frequency)作为对应神经元的链接强度,经过PCNN点火得到源图像在小波域中的点火映射图;通过判决选择算子,选择点火次数多的小波系数作为对应的融合系数,进行区域一致性检验,获到最终的融合系数;对融合后的系数进行小波逆变换得到融合图像。实验结果表明,该方法有效地综合源图像中的重要信息,得到更好视觉效果和更优量化指标的融合图像,在主客观评价上均优于小波、PCNN等方法。For PCNN（Pulse Coupled Neural Network） has particular advantage in image processing,a novel fusion algorithm of multi-sensor image is proposed based on wavelet transform and PCNN.Original images are decomposed by wavelet transform, and the sub-band images at different scales are obtained.A fusion rule is given through making use of synchronous pulse bursts.This method uses SF（Spatial Frequency） of wavelet coefficients at different scales as the linking strength of the corresponding neuron.After the processing of PCNN with the adaptive strength, new fire mapping images in wavelet domain are obtained.According to the fire mapping images,the fusion coefficients are decided by the compare-select operator.The region consistency test is used on the fusion coefficients to obtain the final fusion coefficients.Fusion images are obtained by wavelet inverse transform.Experimental results illustrate that this algorithm is efficient to integrate important information from the original images and obtains fusion images,which outperforms wavelet,PCNN,and so on both in visual quality and objective evaluation index.

关 键 词：图像融合 脉冲耦合神经网络 小波变换 空间频率 点火映射图 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]