摘要
提取源图像的亮度分量,采用非下采样Contourlet变换对其进行分解,通过"合成图像像素值取大"准则对高频系数进行处理得到融合决策图,并对其进行一致性校验,最后根据校验后的决策图在RGB空间进行像素点选取,得到融合图像。实验结果表明,本文方法解决了RGB空间融合方法容易导致的颜色失真,同时本方法仅对亮度分量融合,降低运算复杂度。融合图像在保留图像有用信息的同时,弥补了传统空间域方法在细节表现力上的不足,更加符合人类的视觉特征。本文融合方法还用于灰度多聚焦图像融合中,实践证明,融合效果很好。
The luminance components of the input color images are processed with nonsubsam- pled Contourlet transform （NSCT）. The decision map is attained by fusing the high frequency coefficients through the scheme of synthesis image coefficient value select maximum （SI- CVSM）. Then the decision map is verified by consistency verification and used to select the RGB components simultaneously. Experiment results show that the proposed algorithm avoids color distortion caused by fusion algorithms in RGB space model. At the same time, the meth- od is only the luminance component integration, thereby reducing the computational complexi- ty. The fused image, which retains the important information existing in the input color ima- ges, and improves the detail expression, is more adaptive to the vision of human beings. The integration method is also applied to gray-scale multi-focus image integration, and the practice shows that it is pretty effective.
出处
《数据采集与处理》
CSCD
北大核心
2013年第6期755-760,共6页
Journal of Data Acquisition and Processing
