基于非下采样Bandelet变换的多聚焦图像融合  

Multifocus image fusion based on nonsubsampled Bandelet transform

作　　者：杨扬[1,2] 戴明[1] 周箩鱼[1,2] 孙明超[1] 

机构地区：[1]中国科学院长春光学精密机械与物理研究所,长春130033 [2]中国科学院大学,北京100049

出　　处：《吉林大学学报（工学版）》2014年第2期525-530,共6页Journal of Jilin University:Engineering and Technology Edition

基　　金：'973'国家重点基础研究发展计划项目(2009CB72400102A)

摘　　要：针对同一场景的多聚焦图像,提出了基于非下采样Bandelet变换(NSBT)的图像融合方法。首先求出源图像的平均图像,再利用该图像的小波系数计算出最优的几何流,进而构造出一组Bandelet基,然后将源图像通过这组基进行表示,对得到的系数按所对应的不同物理意义分别进行融合,最后将融合系数进行非下采样Bandelet逆变换得到融合图像。实验结果表明,与基于小波变换(WT)和非下采样小波变换(NSWT)的融合方法相比,本文所提方法取得了更好的视觉效果和量化指标。A novel fusion algorithm based on nonsubsampled Bandelet transform is proposed for the multifocus images of same scene. First, the average image is calculated from the source images, after that the best geometric flow is computed using the wavelet coefficients of the image and a Bandelet basis is constructed. Secondly, the source images are represented using the Bandelet basis, and then the obtained coefficients are fused respectively according to their physical significance. Finally, the fused coefficients are applied on the inverse nonsubsampled Bandelet transform to obtain the fused image. The experiment results indicate that, compared with the fusion methods based on Wavelet transform and nonsubsampled Wavelet transform, respectively, the proposed method can provide better fusion quality in terms of both visual sense and quantified measure.

关 键 词：信息处理技术 图像融合 非下采样Bandelet变换 多尺度分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]