6通道MSVD构造及其在多聚焦图像融合中的应用

Construction of the six channel multi-scale singular value decomposition and its application in multi-focus image fusion

针对经典的奇异值分解(singular value decomposition,SVD)在图像处理中的不足,提出了一种6通道多尺度奇异值分解(multi-scale SVD,MSVD)的构造方法,并将其应用于多聚焦图像融合中。首先,在经典SVD的基础上,利用矩阵分块的方法,给出了一种6通道MSVD的构造方法。其次,对参加融合的多聚焦图像进行6通道MSVD分解,得到高层低频和各层5个方向的高频,对分解的低频子图像采用取平均、高频子图像采用区域能量取大的融合规则进行融合,并进行MSVD逆变换得到融合结果图像。最后,对融合结果图像进行主观分析和客观评价。实验结果表明该方法有好的视觉效果,融合结果图像有较高的清晰度和较丰富的边缘细节信息,且没有方块效应。从客观指标看,该方法有较高的清晰度和空间频率,其清晰度和空间频率比基于离散小波变换、基于提升小波变换、基于曲波变换和基于轮廓波变换的融合方法都高。

In order to solve the deficiencies of classical singular value decomposition （SVD） in image processing, a construction method of the six channel multi-scale singular value decomposition is presented. An image fusion method based on the multi-scale SVD （MSVD） is proposed. Firstly, based on the principle of classical SVD and blocking algorithm, the six-channel MSVD is performed. Secondly, the images involved in the fusion are decomposed into one approximation and five detail images with different resolution by the MSVD. The fusion rule is that the average value is selected for low-frequency sub-image; while for the high-frequency subimages, the coefficients with larger area energy value are employed. The fused image is obtained by using reconstruction method of the MSVD. Finally, the fusion performance of the result image is evaluated using subjective analysis and objective indices. The experimental results show that the proposed fusion method has good visual effect and has no blocking-artifact in the fused images. When compared with the fusion method based on discrete wavelet transform, lifting wavelet transform, curvelet transform and contourlet transform, the proposed method has been observed to have higher definition and spatial frequency.