摘要
针对小波变换容易造成细节信息丢失、非下采样轮廓波变换(NSCT)分解的低频子带系数不稀疏以及红外与可见光图像融合结果综合性能不佳的问题,提出了一种基于稀疏表示和NSCT-PCNN的红外与可见光图像融合算法。首先将源图像进行NSCT分解,获得低、高频子带;其次,利用K奇异值分解(K-SVD)算法对低频子带进行字典训练,实现低频子带的稀疏表示和低频稀疏系数的融合;然后,利用高频子带的空间频率激励脉冲耦合神经网络(PCNN),选择较大点火次数的系数作为高频子带的融合系数;最后对低、高频子带融合系数进行NSCT逆变换,得到融合的图像。实验结果表明,该算法在视觉效果和客观指标方面均具有较大优势,且融合结果综合性能优于现有算法。
In view of the problems of the loss of detailed information caused by wavelet transform,the nonsparsity of low-frequency subband coefficient decomposed by Non-Subsampled Contourlet Transform( NSCT),and the poor comprehensive performance of infrared and visible image fusion,an algorithm for fusion of infrared and visible images is proposed based on sparse representation,NSCT,and Pulse Coupled Neural Network( PCNN). Firstly,the original image is decomposed by NSCT to obtain the low-frequency and high-frequency subbands. Secondly,the K-SVD( Singular Value Decomposition) algorithm is used to carry out dictionary training on the low-frequency subband to realize the sparse representation of lowfrequency subband and the fusion of low-frequency sparse coefficients. Then,the spatial frequency of the high-frequency subband is utilized to stimulate PCNN,and the coefficient with more ignition times is selected as the fusion coefficient of the high-frequency subband. Finally,the NSCT inverse transform is applied to the low and high frequency subband fusion coefficients to obtain the fused image. The experimental results show that the proposed algorithm has a great advantage in subjective visual effect and objective index evaluation,and its comprehensive performance is superior to that of the existing algorithm.
引文
[1]张莹,李言俊,张科,等.基于NSCT的红外与可见光图像融合[J].计算机工程与应用,2011,47(3):196-198.
[2]赵飞翔,陶忠祥.基于NSCT的红外与可见光图像融合[J].电光与控制,2013,20(9):29-33.
[3]陈木生,蔡植善.基于NSCT的红外与可见光图像融合方法研究[J].激光与光电子学进展,2015,52:061002-1-061002-6.
[4]首照宇,胡蓉,欧阳宁,等.基于多尺度稀疏表示的图像融合方法[J].计算机工程与设计,2015,36(1):232-235.
[5]刘雯敏,陈秀宏.基于PCNN和非线性滤波万有引力的医学图像融合[J].计算机工程与应用,2014,50(24):191-198.
[6]张强,郭宝龙.一种基于非下采样Contourlet变换红外图像与可见光图像融合算法[J].红外与毫米波学报,2007,26(6):476-480.
[7]GONG J M,WANG B B,QIAO L.Image fusion method based on improved NSCT transform and PCNN model[C]//IEEE Conference Publications on ISCID,2016:28-31.
[8]张生伟,李伟,赵雪景.一种基于稀疏表示的可见光与红外图像融合方法[J].电光与控制,2017,24(6):47-52.
[9]欧阳宁,郑雪英,袁华.基于NSCT和稀疏表示的多聚焦图像融合[J].计算机工程与设计,2017,38(1):177-182.
[10]赵景朝,曲仕茹.基于Curvelet变换与自适应PCNN的红外与可见光图像融合[J].西北工业大学学报,2011,29(6):849-853.
[11]赵海峰,鲁毓苗,陆明,等.基于快速稀疏表示的医学图像压缩[J].计算机工程,2014,40(4):233-236.
[12]QU G H,ZHANG D L,YAN P F.Information measure for performance of image fusion[J].IET Journals and Magazines on Electronics Letters,2002,38(7):313-315.
[13]PIELLA G,HEIJMANS H.A new quality metric for image fusion[C]//IEEE Conference Publications on Image Processing,2003:173-176.
[14]LIU Z,BLASCH E,XUE Z Y.Objective assessment of multi-resolution image fusion algorithms for context enhancement in night vision:a comparative study[J].IEEE Transactions on PAMI,2012,34(1):94-109.
[15]张小利,李雄飞,李军.融合图像质量评价指标的相关性分析及性能评估[J].自动化学报,2014,40(2):306-315.
[16]郭明,符拯,奚晓梁.基于局部能量的NSCT域红外与可见光图像融合算法[J].红外与激光工程,2012,41(8):2229-2235.
[17]陈浩,朱娟,刘艳滢,等.利用脉冲耦合神经网络的图像融合[J].光学精密工程,2010,18(4):995-1001.
[18]郝文超,贾年.NSCT域内基于自适应PCNN的红外与可见光图像融合方法[J].西华大学学报:自然科学版,2014,33(3):11-14.
[19]LI S,YANG B,HU J.Performance comparison of different multi-resolution transforms for image fusion[J].Information Fusion,2011,12(2):74-84.
[20]SHABANZADE F,GHASSEMIAN H.Multimodal image fusion via sparse representation and clustering-based dictionary learning algorithm in nonsubsampled contourlet domain[C]//IEEE Conference Publications on IST,2016:472-477.