郭凌鑫 袁丽英 王飞越

摘 要：图像融合在军事、医疗等领域有广泛的应用，图像融合技术是指采用特定的算法将两幅或者多幅图像融合成为一幅新的图像。针对离散小波变换图像融合算法中下采样环节在每次滤波后一半的系数被丢弃导致缺乏平移不变性，使得融合算法易受到源图像误配准影响的问题，提出了一种基于非下采样离散小波变换的图像融合算法，采用窗口融合规则，并对融合系数进行一致性检测，弥补下采样环节的不足，提高融合效果。通过对基于离散小波变换和基于非下采样离散小波算法的图像多聚焦融合仿真研究表明，非下采样离散小波变换算法具有平移不变性，而且具有较高融合图像互信息和峰值信噪比，融合效果优于离散小波变换图像融合算法。

关键词：关键字：图像融合；离散小波变换；非下采样离散小波变换

DOI：10.15938/j.jhust.2018.02.004

中图分类号： TN911.73

文献标志码： A

文章编号： 1007-2683（2018）02-0018-05

Abstract：Image fusion is widely used in military， medical and other fields. Image fusion technology is to use a specific algorithm to synthesize two or more images into a new image. Because of down sampling link of discrete wavelet transform image fusion algorithm， half of the coefficients are discarded， resulting in the lack of translation invariance. It makes the fusion algorithm easily affected by the error of source image registration. Due to these shortcomings， an image fusion algorithm based on non-sampling discrete wavelet transform is proposed and window fusion rule is also applied. The coefficient consistency detection makes up for the lack of sampling link and improve the fusion effect. Through discrete wavelet transform and non-sampling discrete wavelet algorithm are applied to image fusion respectively. Non-sampled discrete wavelet transform algorithm has translation invariance and it has high mutual information and peak signal to noise ratio of fusion image. Non-sampled discrete wavelet transform algorithm fusion effect is better than discrete wavelet transform image fusion algorithm.

Keywords：image fusion； discrete wavelet transform； non-subsampled discrete wavelet transform

0 引 言

近二十年来，图像融合技术在军事领域得到了广泛的应用。美军通过微光和多波段红外图像融合方式让其夜视技术处于世界领先的地位[1]。美軍的另一个夜视融合研究项目便是研发多光谱自适应网络战术图像系统（MANTIS），该系统可将可见光、短波和长波红外图像进行融合[2]。此外，美国一些其他的军事工业企业也着力开展基于红外和可见光传感器的夜视图像融合技术[3]。图像融合不仅在军事领域有广泛应用，而且遍及到众多民用应用领域。……