DT-CWT优化边缘特征和自适应PCNN的图像融合  

Optimized edge feature and adaptive PCNN image fusion based on DT-CWT

作　　者：高少忠 沈小林 Gao Shaozhong;Shen Xiaolin(School of Electrical and Control Engineering,North University of China,Taiyuan 030051,China)

机构地区：[1]中北大学电气与控制工程学院

出　　处：《电子测量与仪器学报》2019年第9期114-120,共7页Journal of Electronic Measurement and Instrumentation

摘　　要：针对传统小波变换在图像融合中易丢失细节、忽略边缘特征且清晰度不高的现状,提出双树-复小波变换(DT-CWT)优化边缘特征和自适应脉冲耦合神经网络(PCNN)的图像融合算法。首先将原图经双树复小波处理,并对低频分量和高频分量进行分析。然后低频分量选用边缘特征和清晰度选择融合算法,保留较多细节信息,高频分量以方向信息自适应调整PCNN连接强度,把改进的拉普拉斯能量和作为自适应PCNN网络的输入,将点火输出幅度的总值设为高频分量的系数。最终进行双数复小波逆变换处理。实验结果表明,较已有图像融合算法,该算法融合得到的图像在主客观评价方面都有提高,互信息量(MI)提高了0. 127 4~2. 450 4,边缘保持度(QAB/F)提高了0. 069~0. 256。该算法突出了融合图像的目标信息,极大地保留图像边缘、纹理等有用信息,使图像更清晰。Aiming at the current situation that traditional wavelet transform is easy to lose detail and ignore edge features and low definition in image fusion, this paper proposes optimized edge feature and adaptive PCNN image fusion based on DT-CWT. Firstly, the original image was processed by double tree complex wavelet, and the low-frequency component and high-frequency component were analyzed. Then the low-frequency component used the edge feature and the sharpness selection fusion algorithm to retain more detailed information, the high-frequency component adaptively adjusted the PCNN connection strength with the direction information, and the improved Laplacian energy was used as the input of the adaptive PCNN network. The total value of the ignition output amplitude was set as the coefficient of the high frequency component. Finally, the double complex wavelet inverse transform process was performed. Compared with the existing image fusion algorithm, the experimental results show that the images obtained by the fusion of the algorithm are improved in subjective and objective evaluation, mutual information (MI) is increased by 0. 127 4 ~ 2. 450 4, and edge retention (Q^AB/F)is increased by 0. 069 - 0. 256. The algorithm highlights the target information of the fused image, and greatly preserves useful information such as image edges and textures to make the image clearer.

关 键 词：图像融合 双树-复小波 边缘特征 自适应脉冲耦合神经网络 方向信息 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TN0[自动化与计算机技术—计算机科学与技术]