Pixel-level image fusion based on multi-to-multi turbo iterative
Pixel-level image fusion based on multi-to-multi turbo iterative
复制标题
DOI：
10.1109/apsar.2007.4418729
复制
发表时间：
2007-11
期刊：
影响因子：
--
通讯作者：
Na Li;Chu He;Lei Yu;Hong Sun
中科院分区：
文献类型：
--
作者：
Na Li;Chu He;Lei Yu;Hong Sun
提出了一种基于多对多Turbo迭代的像素级图像融合方法。提出了一种新的多对多图像融合思想。毫无疑问，用一幅图像来覆盖多幅图像中包含的所有信息是不可能的。因此，多对多图像融合系统比多对一图像融合系统更加合理。基于这种思想，获得多幅正交图像将具有很好的应用前景。本文将局部线性生成模型方法扩展到从源图像中提取正交图像。此外，还采用了Turbo迭代的本质，迭代地使用两种正交融合算法。在迭代过程中，来自两个算法的信息被交换。本文迭代地使用了基于小波的方法和基于扩展模型的方法。基于小波变换的融合方法虽然能保留更多的细节信息，但对噪声比较敏感。相比之下，基于模型的融合方法可以很好地控制噪声。该方法同时采用了多对多的思想和Turbo迭代算法，能够得到更好的融合结果，具有更多的细节和更少的噪声。实验结果令人满意。
A pixel-level image fusion method based on multi-to-multi turbo iterative is proposed. An innovative multi-to-multi image fusion idea is introduced in this paper. Undoubtedly, it's impossible to cover all the information contained in several images by one image. So multi-to-multi image fusion system is much more reasonable than multi-to-one image fusion system. Based on this idea, getting several orthogonal images can provide a promising application in future. The local linear generative model method is expanded to get orthogonal images from the source images in this paper. In addition, turbo iterative essence is also adopted; two orthogonal fusion algorithms are used iteratively. The information from two algorithms is exchanged during the iterative process. This paper uses the wavelet based method and the expanded model based method which is mentioned above iteratively. Although the wavelet based fusion method can retain much detailed information, it is sensitive to noise. In contrast, the model based fusion method can control the noise well. With the multi-to-multi idea and turbo iterative being both adopted, the proposed image fusion method can get a better fusion result with more details and less noise. The experiments show a satisfying result.