Abstract: The neighboring relation of low-frequency wavelet coefficients and the directional characteristic of the high-frequency coefficients is discussed and analyzed， then， a novel image fusion algorithm based on directional spread of wavelet coefficients’ differences is proposed. At first， for every coefficient of the high-frequency， the algorithm of this paper makes out the fusing coefficient by using the directional templates to gain the wavelet coefficients’ difference. Then， for every coefficient of the low-frequency， the algorithm of this paper makes out the fusing coefficient by using the variance. At last， the fused image is obtained through the reversely lifting wavelet transformation. Entropy and Average grads are used to judge the results of the proposed algorithms. Experimental results show that the proposed method is more outstanding than the conventional methods， and it has some application value.

Key words: image fusion, lifting wavelet transformation, difference spread, directional characteristic, fusion rule

摘要： 对图像经提升小波变换后低频小波系数的区域相关性以及高频小波系数的方向特性进行了分析和统计，提出了一种基于小波系数差值方向传播的图像融合新方法。首先，对于高频子带的每一个系数，依据其所在子带内系数分布的方向特征，将待融合系数的差值按照一定的方向模板进行加权，通过比较对应位置加权后的数值，进而确定高频融合系数；对低频子带的每个系数，依据其系数的相关性，通过比较其八邻域方差，进而确定低频融合系数，最后，对融合系数进行提升小波逆变换，获得融合后的图像。采用信息熵和清晰度对融合后的图像进行评价，实验结果表明，算法所获得的融合结果明显优于传统的图像融合算法，具有一定的实用性。

关键词: 图像融合, 提升小波变换, 差值传播, 方向特性, 融合规则