应用相关系数的小波变换图像融合算法  

A Wavelet Transform Image Fusion Algorithm Based on Related Coefficients

作　　者：闫利伟[1] 任小康[1] 

机构地区：[1]西北师范大学数学与信息科学学院,甘肃兰州730070

出　　处：《计算机工程与科学》2011年第11期103-107,共5页Computer Engineering & Science

基　　金：甘肃省自然科学基金资助项目(1010RJZA011)

摘　　要：多聚焦图像融合的关键问题是如何保持原始图像的边缘和细节信息,由此本文提出了一种基于相关系数的小波变换图像融合策略。针对小波分解的不同频率域,设计了两种选择高频系数和低频系数的原则。选择高频系数时,采用绝对值最大、平均与选择相结合两个原则;选择低频系数时,基于平均与选择相结合、相关系数两个原则。最后将提出的算法用于多聚焦图像的融合,并利用客观评价指标对该融合方法的性能进行了评价和分析。实验结果表明,本文方法可以更好地保留多幅原图像中的有用信息,得到多个目标聚焦都很清晰的融合图像,并且各项融合评价指标均有所提升,说明本文算法有一定的有效性和实用性。The key problem in multi-focus image fusion is how to keep the detail and edge in the original image. Therefore, a new image fusion algorithm is developed based on the related coefficients of wavelet transform. The schemes of choosing high frequency and low frequency coefficients are two approaches designed. While choosing high frequency coefficients, we select these that have the maximal absolute values and combination of averages with selection. When choosing low frequency coefficients,we select two principles based combination of average with selection and coefficients related. The algorithm is used to fuse two multi-focused images. In addition, the performance of the fusion methods is appraised and analysed with objective evaluation index. The experiments show that this algorithm can preserve all the useful information from the primitive images and all the targets in the fused image are very clear. Each fusion evaluation index achieves improvement. The algorithm is of certain effective and parctial.

关 键 词：图像融合 小波变换 平均与选择 相关系数 融合质量评价 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]