改进快速NSST的热烟羽红外与可见光图像融合  

Infrared and visible images fusion of thermal plume based on rapid non-subsampled shearlet transform

作　　者：方静[1] 罗高鹏[1] 

机构地区：[1]合肥工业大学计算机与信息学院,安徽合肥230009

出　　处：《激光与红外》2017年第7期914-920,共7页Laser & Infrared

基　　金：国家自然科学基金项目(No.61102154)资助

摘　　要：针对非下采样剪切波变换逐步划分子带过程中迭代滤波运算效率不高的问题,提出一种非下采样剪切波变换的快速分解重构方法。改进的分解方法根据分解的尺度卷积非下采样金字塔滤波器组中相应的滤波器,用卷积结果对源图像进行滤波,直接生成子带图像,在提高图像分解速度的同时,也相应减少分解迭代滤波的运算量。本文采用一种基于标准差的区域特性量测方法用于低频系数融合,突出低频系数中的轮廓信息;最高频系数融合采用能量取大的融合规则,其余高频系数通过结合改进的拉普拉斯能量和与梯度特性进行融合,使目标烟羽的边缘信息更加清晰。实验结果表明,融合后热烟羽图像目标细节清晰突出;综合客观评价指标来看,本文方法优于一些常用的红外与可见光图像融合方法。A rapid image decomposition and reconstruction method based on non-subsampled shearlet transform is presented to improve the iterative filter efficiency in the process of dividing the sub-band images. All filters in the non- subsampled pyramid fiher bank are convoluted based on the scales of the decomposition. Source image is filtered with convoluted results to acquire the sub-band images directly,which reduces the computational complexity of iterative filtering. To highlight the contour information, the low frequency coefficients are fused by using regional characteristics measuring method based on the standard deviation. The highest frequency coefficients are fused by using the maximum energy selection rule; the other high frequency coefficients are fused by combining sum modified-Laplacian （SML） with gradient features ,which highlights the edge information of the target plume. Experimental results show that the subjective visual effect of the fused images is well,which can reflect the thermal plume＇ s temperature distribution obviously. Through several objective evaluation indexes ,the proposed fusion method is superior to several traditional infrared and visible images fusion algorithms.

关 键 词：红外图像 图像融合 非下采样剪切波变换 快速分解 边缘特性 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]