一种改进的DCT域压缩图像融合算法  

An Improved Image Fusion Algorithm in DCT Compressed Domain

作　　者：赖思渝[1] 王娟[2] 

机构地区：[1]川北医学院医学影像学院,四川南充637000 [2]西华师范大学计算机学院,四川南充637002

出　　处：《网络新媒体技术》2016年第3期10-16,共7页Network New Media Technology

基　　金：四川省教育厅重点项目"图像融合及关键技术研究"(编号:14ZA0124)

摘　　要：为提高图像融合算法性能,本文提出一种基于对比度的离散余弦图像融合算法,通过将图像划分为8×8的数据块并分别计算其DCT系数。其中,对AC系数采取取大值的方法,DC系数采取取平均值的方法。通过与另两种融合算法的对比实验,提出的算法从主观视觉及客观评价指标上均优于DCT平均法,该算法仅有一项评价指标数值优于小波法,但从视觉融合效果上看已完全达到小波法的视觉标准,而且通过对10组融合图像的耗时统计发现,提出的算法与DCT平均法耗时相近且远低于小波法。当生成的融合图像需要以JPEG格式存储或传输,或待融合图像为JPEG格式时,提出算法比小波法更简便且具有更低的计算复杂性。In order to improve the performance of image fusion algorithm, a discrete cosine image fusion algorithm based on contrast is proposed in the paper. In the algorithm the source image is divided into the blocks of 8 by 8 and the DCT coefficient for each block is calculated respectively. In this paper, the max value method is adopted to calculate the AC coefficient while the averaging value method is took in DC coefficient calculation. By comparison with the other two algorithms, the proposed algorithm is superior to the DCT average method from subjective and objective aspects. And the algorithm has one index outperform that of wavelet method, but gets nearly the same visually standard. In the 10 groups of fused images time consumption statistics experiment, the proposed algorithm and the DCT average method are similar in time cost and the former is more time saving when compare to wavelet. When the fused image needs to be stored or transmitted in JPEG format or the source image is in JPEG format, the proposed algorithm is simpler and has a lower computational complexity than wavelet.

关 键 词：图像融合 离散余弦变换 小波变换 对比度 联合图像专家组 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]