基于区域灰度统计信号处理的图像融合方法

Region-based statistical signal processing scheme for image fusion

针对多源图像融合问题,提出了一种在多分辨率框架下基于区域内灰度特征统计信号的融合算法.利用图像灰度特征的区域生长法对源图像进行区域分割,并以裂缝边缘作为特征区域的闭合边界,对源图像与分割结果的区域映射图作多分辨率变换.在图像低频部分,以联合区域映射图为指导,在区域内建立信号与噪声的高斯混合分布模型,利用期望极大化(EM,Expectation Maximization)算法迭代估计噪声模型分布参数,获得低频融合结果;在图像高频部分,根据系数在区域映射图上的位置差异分别采用窗口系数加权平均法和系数绝对值选大法进行融合,将低频和高频融合结果反变换得到最终融合图像.融合结果表明:该方法是可行和高效的,且比其他图像融合方法具有更好的性能.

A new image fusion scheme based on region statistical signal processing was proposed. The region growing technique using gray-level clustering was employed to segment the source images into different regions whose borderline represented with crack edge. The registered source images and their segmented mapping were decomposed into a multi-resolution representation with both low-frequency coarse information and high-frequency detail information respectively. The expectation maximization algorithm modeled with noise statistic distribution was used to fuse the low-frequency coarse information of the registered images, while the match and salience measures were applied to fuse the high-frequency detail information of the registered images. The final fused image was obtained by taking the inverse transform of the composite multi-resolution representations information. Fusion experiments on real world images indicate that the proposed method is effective and efficient, which achieves better performance than the most generic fusion method.