Abstract: A new multi-resolution analysis fusion algorithm introducing directional contrast and region window standard deviation is proposed, which considers the correlation between approximation and details. Weighted factors are adaptively adjusted to gain fused low frequency parts, meanwhile, directional  contrast maximum and local small window standard deviation maximum are respectively adopted to form fused high frequency parts from decomposed details. Experiments show that fused image using the new algorithm considering the correlation between decomposed low frequency and high frequency coefficients is prior to that of other algorithms focusing on merging these coefficient separately regardless of visual effect or objective assessment metrics. In contrast, the fused image by the new algorithm is rich in information, features are clear, greatly enhancing contrast, improving spatial resolution and retaining spectral information of original images. Obviously, it is feasible and effective.

Key words: wavelet decomposition, multi-resolution analysis, high frequency coefficient, low frequency coefficient, directional contrast, region standard deviation

摘要： 小波变换的图像融合方法已成为现今研究的一个热点。但几乎所有的算法都是在小波域不同尺度上分别对高频系数和低频系数进行融合，没有考虑到它们之间固有的相关性。为此，提出了一种基于方向对比度和区域标准差最大的融合新算法，主要特点是在低频部分采用加权因子自适应调节参数融合，以减少边缘模糊，对于高频部分采用方向对比度和局部区域窗口标准差最大值作为高频分量，突出对比度和局域细节，实验表明：融合后影像信息量丰富，地物轮廓清晰可辨，对比度大大加强，空间分辨率得到了提高，最大限度保留了原始影像的光谱信息，是一种可行有效的融合方法。

关键词: 小波分解, 多分辨率分析, 高频系数, 低频系数, 方向对比度, 区域标准差