基于源图像分析的动态ICA域灰度图像融合方法  

Dynamic ICA Domain Grayscale Image Fusion Method Based on the Source Image Analysis

作　　者：骆媛[1] 金伟其[1] 王岭雪[1] 高绍姝[1] 

机构地区：[1]北京理工大学光电学院,北京100081

出　　处：《北京理工大学学报》2014年第4期407-411,共5页Transactions of Beijing Institute of Technology

基　　金：国家自然科学基金重点资助项目(61231014);十二五总装重点预研项目(40405030302)

摘　　要：提出一种利用源图像之间的冗余、互补等相关信息的动态ICA域融合方法.针对ICA变换函数训练以及ICA域融合规则设计对算法适应性、计算量、图像融合效果和人眼视觉系统识别能力等的影响,研究提出充分利用源图像的冗余和互补信息,从源图像中自动选择有代表性的图像块作为训练图像块,实现ICA域变换函数随源图像而动态变化.将文中方法与N.M 2007方法及双树复小波融合算法作为对比,通过人眼主观感知和客观融合评价参数对融合图像进行对比分析,结果表明,文中方法增强了算法的适应性并降低了计算量,增强了图像融合效果,建立了对场景更全面和精确的表达,可使观察者更有效获得场景信息.Proposed a dynamic independent component analysis domain fusion scheme based on the redundancy and complementary information of the source image analysis. Based on analyzing the traditional ICA domain grayscale image fusion algorithm, the dynamic ICA method was proposed that choosing the specific image blocks from the source images as training image blocks, which made full use of the redundancy and the complementary information of the source images without considering the type of the image and the ICA domain transform function changes with the source images dynamically. The proposed method was compared with N. M 2007 method and dual tree complex wavelet transform fusion method. The simulation result shows that the proposed method enhances environment applicability, reduces computation, establishes a more comprehensive and accurate expression of the scene information, further improves scene perception.

关 键 词：图像融合 独立成分分析 局域图像块 双树复小波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]