提出了一种综合纹理和边缘信息的多尺度图像融合方法。该方法通过高斯滤波获得原始图像的金字塔分解,考虑高斯滤波器和Laws纹理提取滤波器以及边缘梯度滤波器之间的线性关系,采用奇异值分解法求出纹理和边缘图像的对应系数;利用各尺度图像的这些特征,对分解后图像的各层进行表示,并采用基于相似性测度和显著性测度的融合策略进行融合。为了衡量融合结果,采用了一个客观图像融合性能评价测度对融合结果进行评价。最后用若干组图像数据对该算法进行了仿真。试验表明,该方法比传统的梯度金字塔分解和基于纹理的金字塔分解方法具有更好的融合效果。

A newmulti-scale scheme for image fusion is presented using texture and edge information in this paper.The origi-nal images are decomposed into texture and edge maps by pyramid structures.The coefficients corresponding to the filtered tex-ture and edge maps are solved by the singular value decomposition and the linear relationship between Gauss filter and Laws tex-ture extract filter.The decomposed images are represented using the texture and edge features in each scale,and fused using the match and salience measures based fusion strategy.To evaluate the fusion results,objective image fusion performance metrics are used.The experiment results demonstrate by several real image date that the proposed scheme can achieve better fusion perfor-mance than the traditional pyamid decomposition methods.