Abstract: The construction of NonSubsampled Directional Filter Bank（NSDFB） and àtrous wavelet transform which is full shift invariant are introduced，and a novel multi-sensor image fusion scheme based on NSDFB combining àtrous wavelet transform for multi sensor images is proposed.Firstly，each source image is decomposed in highpass subbands and lowpass subbands by àtrous wavelet transform，and the highpass subbands are decomposed in multi-directional high pass-bands respectively by NSDFB.Then different fusion rules are used in different subbands，where local gradient preferential image fusion rule is adopted in highpass subbands and average image fusion rule is adopted in lowpass subbandds.Finally the fused image is reconstructed by taking inverse NSDFB transform and inverse àtrous wavelet transform.The experimental results show that the performance of the novel algorithm is better than the traditional fusion algorithms based on wavelet transformation，and especially，the novel fusion algorithm can effectively eliminate the spectrum warping and the artificial edges caused by wavelet transform.

Key words: image fusion, nonsubsampled directional filter bank, àtrous wavelet transform, shift invariant

摘要： 分析和研究了非下采样方向滤波器组及具有平移不变性的àtrous小波变换的图像变换的优点，提出了一种基于多方向àtrous小波变换的图像融合方法。首先利用àtrous小波变换将待融合源图像分解成不同尺度，不同分辨率的高低频分量，再对高频分量利用非下采样方向滤波器组进行方向分解，然后采取不同的融合方法对分解的高低频分量进行融合处理，低频系数采取平均加权法融合，高频系数则采取局部梯度优先的加权法融合，最后将融合的各频带进行逆非下采样方向滤波器组变换和逆àtrous小波变换得到融合图像。实验表明，在几种不同的客观评价标准下，该方法优于传统小波域中的融合效果，能有效地消除小波变换所带来的光谱扭曲和假边缘现象。

关键词: 图像融合, 非下采样方向滤波器组, àtrous小波变换, 平移不变性