结合区域特性的非下采样剪切波图像融合  

Image fusion based on area characteristics in finite non-subsample shearlet domain

作　　者：周岩[1] 王雪瑞[1] ZHOU Yan;WANG Xue-rui(College of Computer, Henan Institute of Engineering, Zhengzhou 451191,China)

机构地区：[1]河南工程学院计算机学院,河南郑州451191

出　　处：《计算机工程与设计》2017年第2期460-464,477,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(61301232;U1304608);河南省基础与前沿技术研究基金项目(122300410061)

摘　　要：为在图像融合过程中更好地提取图像的有用信息,结合非下采样剪切变换(NSST)的方向敏感性和平移不变性,提出一种结合区域特性的非下采样剪切波图像融合算法(NSST_ISC)。对经过NSST分解得到的低频子带系数采用区域四分位值与区域空间频率相结合的融合算法,高频选用区域能量对比度的融合方案;应用非下采样剪切波逆变换重构得到融合图像,对视觉效果和客观指标分别进行评价。对多聚焦图、红外与可见光图和遥感图进行融合实验,实验结果表明,NSST与融合策略相结合得到的融合图像在主观视觉上优于其它算法,熵值、互信息量和边缘相似度这3项客观指标有较大提高。To better extract the useful information of the image in the process of image fusion,combining with directional sensi-tivity and shift invariance of non-subsample shearlet transform,an image fusion algorithm based on area characteristics in the do-main of non-subsample shearlet transform was proposed.The fusion principle of low frequency sub-band coefficients decomposed using NSST was based on the method of regional-interquartile and regional spatial frequency.As to the high frequency sub-band coefficients,combining regional energy and contrast was adopted as the fusion rule.The low-frequency coefficients and high-fre-quency coefficients by processing were reconstructed to image using the proposed method,both subjective visual evaluation and objective performance assessments of the fusion results were implemented.The fusion experiment based on multi focus image,infrared and visible image and remote sensing image was carried out,the results show that the fusion images of the fusion strate-gy combined with NSST are not only superior to that of other fusion algorithms on subjective evaluation,but on the objective pa-rameters such as entropy,mutual information and edge similar degree increase as well.

关 键 词：图像融合 非下采样剪切波 平均梯度 区域能量 平移不变性 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]