基于NSST图像特征的多聚焦图像融合方法  

Fusion Technique for Multi-focus Images Based on Non-subsampled Shearlet Transform and Image Characteristics

作　　者：张耀军[1] 栗磊[1] 吴桂玲[1] 

机构地区：[1]信阳农林学院计算机系,河南信阳464000

出　　处：《电视技术》2015年第15期129-135,共7页Video Engineering

基　　金：河南省基础与前沿技术研究计划项目(132300410452)

摘　　要：针对广泛存在于多聚焦图像融合方法中的局部图像细节不清晰的现状,提出了一种基于非下采样剪切波变换(NSST)图像特征的多聚焦图像融合方法。利用NSST对待融合源图像进行多尺度、多方向稀疏分解,分别获取低频和一系列高频子带图像。通过空间频率和局部能量确定融合后的低频子带系数,利用边缘检测算子直接获取高频子带图像中的细节和边缘信息,并采取NSST反变换得到最终融合结果图像。仿真实验结果表明,同现有的几种经典算法相比,提出的方法获得的结果图像拥有更清晰的视觉效果、更理想的客观指标效果以及更高的算法运行效率。Faced with the current situation that the visual information of local image details is blurred in many fusion techniques for multi-focus images, a novel fusion technique for multi-focus images based on non-subsampled shearlet transform （NSST） and i,n- age charaeterislies is proposed. The low-frequeney and s series of high-frequency sub-band images are obtained by conducting the multi-scale and multi-directional decompositions of the source images via NSST. The spatial frequeney and local energy are uti- lized to determine the fused coefficients of low-frequeney sub-band. The information of the details and edges existing in the high- frequency sub-band images can be directly aehieved by the edge detection operator. The final fused image can be obtained via in- verse NSST. Simulation experimental results demonstrate that, compared with several current typical algorithms, the fused image based on the proposed technique has much clearer visual performanee, more reasonable objective metric results and much higher functional efficienev.

关 键 词：非下采样剪切波变换 空间频率:局部能量 边缘检测 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程]