树状小波局部对比度与图像融合及性能评估  

Image fusion and performance evaluation based on tree-structured wavelet local contrast

作　　者：同武勤[1] 凌永顺[1] 杨华[1] 马东辉[2] 朱斌[2] 

机构地区：[1]解放军电子工程学院安徽省红外与低温等离子体重点实验室 [2]解放军电子工程学院光电系,合肥230037

出　　处：《计算机工程与应用》2007年第24期230-232,共3页Computer Engineering and Applications

摘　　要：针对红外热成像中目标识别和跟踪的特点,提出了一种基于树状小波变换的局部对比度融合算法,首先采用多尺度树状小波变换的方法对已配准的源图像在相应的能量准则下进行分解,克服了小波变换的移变性;对于分解后的低频子图像采用加权平均的融合规则,对于高频子图像采用局部对比度量测的融合规则,融合图像既保持了源图像的细节信息,又滤去了红外成像中的各类噪声。采用交叉分辨力评价算子量测红外图像中目标与背景的衬比度;通过仿真及基于客观的图像融合评价标准,分别从信息熵、标准差、平均梯度和交叉熵四个参数对图像融合效果进行评估,证明了对于双波段红外辐射图像的融合,提出的融合算法优于小波融合算法、形态学区域分割融合算法。算法尤其适用于红外热成像系统的目标识别和跟踪。According to the feature of target recognition and track in infrared thermal imaging,a new local contrast fusion algorithm is put forward based on tree-structured wavelet transform.The source images are decomposed by dual scale tree-structured wavelet transform under the relevant energy rule.The tree-structured wavelet gets over the move-change character of wavelet.The weighting and mean fusion rule is used to the low frequency images and the local contrast fusion rule is used to high frequency image.The fusion image both keeps the detail information and all kinds of noise are denoised.The across resolution evaluation operator of infrared target and background is presented to measure the contrast.Under the simulation and external evaluation criterion of image fusion,the fusion image is evaluated by information entropy,criterion difference,mean grads and cross entropy.The experiment results show this paper algorithm is better than the wavelet fusion algorithm and morphology region segmentation fusion algorithm and proves this paper algorithm has a better use in the target recognition and track of infrared thermal imaging system.

关 键 词：双波段红外图像树状小波图像融合交叉分辨力 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]